WEBVTT

1
00:00:00.000 --> 00:00:01.760
has his AI joining.

2
00:00:01.760 --> 00:00:03.480
That's funny.

3
00:00:03.480 --> 00:00:05.040
But Janata is not here today.

4
00:00:05.040 --> 00:00:10.320
And we do have one new person, Ira, who's here.

5
00:00:10.320 --> 00:00:15.000
So he's kind of taking Janata's place for the live call here.

6
00:00:15.000 --> 00:00:17.040
He just joined two weeks ago.

7
00:00:17.040 --> 00:00:19.960
So we're doing a trial project right now.

8
00:00:19.960 --> 00:00:25.240
And we've been fast-tracking him to get caught up on ShapeUp.

9
00:00:25.240 --> 00:00:27.000
But so far, he's been crushing it.

10
00:00:27.000 --> 00:00:28.160
He's been doing a great job.

11
00:00:28.160 --> 00:00:31.720
So we just wanted to hear it direct from here

12
00:00:31.720 --> 00:00:32.920
as we go through it.

13
00:00:32.920 --> 00:00:34.880
Yeah, and no better way to get caught up

14
00:00:34.880 --> 00:00:36.800
than to just dive into the middle of a session

15
00:00:36.800 --> 00:00:37.460
together, right?

16
00:00:37.460 --> 00:00:39.520
So that's awesome.

17
00:00:39.520 --> 00:00:40.040
Cool.

18
00:00:40.040 --> 00:00:40.360
All right.

19
00:00:40.360 --> 00:00:41.440
And hey, Lucian and Chris.

20
00:00:41.440 --> 00:00:44.040
Nice to see you guys again.

21
00:00:44.040 --> 00:00:45.520
Good to see you, too.

22
00:00:45.520 --> 00:00:50.440
So OK, so Bruce, do you want to start us off

23
00:00:50.440 --> 00:00:52.640
on what are you guys working on?

24
00:00:52.640 --> 00:00:57.160
And then lead into what do you have so far?

25
00:00:57.160 --> 00:01:00.280
And then we'll do like before, and we'll jump straight in.

26
00:01:00.280 --> 00:01:01.120
Yeah, this is great.

27
00:01:01.120 --> 00:01:02.720
I think so we have a couple of things.

28
00:01:02.720 --> 00:01:07.120
And we even had a call to try to frame up this call

29
00:01:07.120 --> 00:01:11.360
to figure out what we do to get the most out of it.

30
00:01:11.360 --> 00:01:16.560
I think we definitely have one project we could go deep on.

31
00:01:16.560 --> 00:01:17.760
It's not related to Tribe.

32
00:01:17.760 --> 00:01:19.200
It's another client's project.

33
00:01:19.200 --> 00:01:23.280
But it's got some very unique rabbit holes and things

34
00:01:23.280 --> 00:01:26.000
that some new tech that we've not worked with before.

35
00:01:26.040 --> 00:01:28.760
So there's a lot of unknowns.

36
00:01:28.760 --> 00:01:32.080
And I thought one way to get to that project

37
00:01:32.080 --> 00:01:36.520
is just talk about the process of how we worked with the client

38
00:01:36.520 --> 00:01:39.760
and how we got to the point of having this project.

39
00:01:39.760 --> 00:01:41.160
Because I think there's some work,

40
00:01:41.160 --> 00:01:44.160
there's some pre-work that could have been done potentially

41
00:01:44.160 --> 00:01:46.640
weeks ago that we've inadvertently

42
00:01:46.640 --> 00:01:48.280
left to inside the project.

43
00:01:48.280 --> 00:01:50.760
So now the clock started, so to speak.

44
00:01:50.760 --> 00:01:55.560
And we're now feel like we're, I mean,

45
00:01:55.560 --> 00:01:58.280
I would probably guess we're like maybe 40%, 50% shaped.

46
00:01:58.280 --> 00:02:04.880
60% is probably a lot less, like it's less optimistic.

47
00:02:04.880 --> 00:02:07.200
So we're trying to figure out, because our process,

48
00:02:07.200 --> 00:02:09.120
as you know, is a little bit unique in that we

49
00:02:09.120 --> 00:02:10.919
have the client relationship.

50
00:02:10.919 --> 00:02:13.680
I'll say one thing that has worked extremely well

51
00:02:13.680 --> 00:02:15.800
is I have, on two different clients,

52
00:02:15.800 --> 00:02:19.520
I have just pulled them right into the entire process.

53
00:02:19.880 --> 00:02:22.320
I've gotten on a three-hour call with some of them,

54
00:02:22.320 --> 00:02:25.560
four-hour call, and we breadboard, we shape,

55
00:02:25.560 --> 00:02:27.240
we set the appetite.

56
00:02:27.240 --> 00:02:28.960
They see the whole process.

57
00:02:28.960 --> 00:02:32.240
And that has been incredibly eye-opening for them

58
00:02:32.240 --> 00:02:34.240
and for us to get super clear.

59
00:02:34.240 --> 00:02:36.400
And so we're working more and more that way

60
00:02:36.400 --> 00:02:38.440
where we just bring them into the team.

61
00:02:38.440 --> 00:02:41.040
Because there's not a way to sort of make it all packaged

62
00:02:41.040 --> 00:02:41.840
and then just show them.

63
00:02:41.840 --> 00:02:44.240
You kind of have to bring them to the desk.

64
00:02:44.240 --> 00:02:47.400
Because they are kind of the key decision maker,

65
00:02:47.400 --> 00:02:49.720
especially when it comes to scope and trade-offs

66
00:02:49.720 --> 00:02:52.240
and out-of-bounds stuff.

67
00:02:52.240 --> 00:02:56.480
So that's where we're finding a lot of value.

68
00:02:56.480 --> 00:03:00.640
So already, I mean, this is only our second month,

69
00:03:00.640 --> 00:03:05.240
third call, like crazy amount of value already.

70
00:03:05.240 --> 00:03:06.200
Awesome.

71
00:03:06.200 --> 00:03:09.560
Like worth all of our time and energy to do it.

72
00:03:10.840 --> 00:03:13.080
And so what I'm seeing now,

73
00:03:13.080 --> 00:03:14.640
now that we've kind of got a good process,

74
00:03:14.640 --> 00:03:16.720
I'm trying to make sure that we sort of prep

75
00:03:16.720 --> 00:03:19.960
the process that happens before this.

76
00:03:19.960 --> 00:03:22.560
And how much do we frame at a time?

77
00:03:22.560 --> 00:03:24.600
Because at some point we kind of say,

78
00:03:24.600 --> 00:03:25.760
okay, this is four weeks.

79
00:03:25.760 --> 00:03:27.880
And then the clients lock in and they commit to it.

80
00:03:27.880 --> 00:03:30.240
And we have an agreement that says, this is four weeks.

81
00:03:30.240 --> 00:03:32.480
Here's the dollars that have to be paid.

82
00:03:32.480 --> 00:03:33.920
And then we lock in at four weeks.

83
00:03:33.920 --> 00:03:37.160
And so then we now need to go like,

84
00:03:37.160 --> 00:03:39.280
we want to try to keep to a four-week calendar

85
00:03:39.280 --> 00:03:41.640
as well as a four-week scope as well.

86
00:03:42.520 --> 00:03:46.360
Because if we run over in calendar or budget,

87
00:03:46.360 --> 00:03:48.920
like in hours, it costs us.

88
00:03:48.920 --> 00:03:50.720
And not just in the team,

89
00:03:50.720 --> 00:03:52.000
because now we can't like,

90
00:03:52.000 --> 00:03:53.320
other projects get pushed back.

91
00:03:53.320 --> 00:03:55.960
We get over, you know, people start getting overwhelmed.

92
00:03:55.960 --> 00:03:56.960
Totally.

93
00:03:56.960 --> 00:03:59.000
But also then, you know, we just want to be as,

94
00:03:59.000 --> 00:03:59.840
that's part of it.

95
00:03:59.840 --> 00:04:01.600
There's a healthy,

96
00:04:03.720 --> 00:04:05.760
one of the healthy boundaries is that timeline.

97
00:04:05.760 --> 00:04:08.360
Like you have the four weeks, we're on week two.

98
00:04:08.360 --> 00:04:10.280
Like we only have two weeks, like, you know,

99
00:04:10.280 --> 00:04:12.080
like we have to make some trade-offs now

100
00:04:12.080 --> 00:04:13.320
to get this thing shipped

101
00:04:14.600 --> 00:04:16.200
instead of just letting it, you know,

102
00:04:16.200 --> 00:04:17.120
get into scope creep.

103
00:04:17.120 --> 00:04:20.480
And then, so I can, if it's helpful,

104
00:04:20.480 --> 00:04:21.880
I made a little diagram of just kind of

105
00:04:21.880 --> 00:04:23.360
how our process is working.

106
00:04:23.360 --> 00:04:25.720
I can just, I just sketched it out on a little whiteboard,

107
00:04:25.720 --> 00:04:27.120
but I can show it to you.

108
00:04:27.120 --> 00:04:29.000
And then I can show you how we got to this project

109
00:04:29.000 --> 00:04:30.120
or we can pause for a second

110
00:04:30.120 --> 00:04:32.000
if you have any questions so far.

111
00:04:32.000 --> 00:04:33.920
So, so let's see.

112
00:04:33.920 --> 00:04:36.480
So what we're trying to get to here is,

113
00:04:37.360 --> 00:04:42.000
you're trying to work out what the process looks like

114
00:04:42.000 --> 00:04:43.720
of like, what do you do with the client

115
00:04:43.720 --> 00:04:46.240
before this four-week clock starts?

116
00:04:46.240 --> 00:04:47.520
And what do you do with them

117
00:04:47.520 --> 00:04:48.840
after the four-week clock starts?

118
00:04:48.840 --> 00:04:49.840
That kind of the question is like,

119
00:04:49.840 --> 00:04:52.200
what kind of work is happening at which stage

120
00:04:52.200 --> 00:04:54.120
in that life cycle?

121
00:04:55.440 --> 00:04:56.280
Right.

122
00:04:56.280 --> 00:04:57.120
And figuring out,

123
00:04:58.600 --> 00:05:00.360
so we typically do an engagement.

124
00:05:00.000 --> 00:05:06.120
That's kind of a one, one or two week, uh, process where we just connect with them.

125
00:05:06.160 --> 00:05:07.560
We figure out the lay of the land.

126
00:05:07.560 --> 00:05:08.500
What is the project?

127
00:05:08.500 --> 00:05:13.600
We kind of identify two or three potential pitches.

128
00:05:13.600 --> 00:05:17.840
There'll be like, we just frame two or three potential projects and say, Hey,

129
00:05:17.840 --> 00:05:20.840
we could, we could work with you if you want, these are the three next things.

130
00:05:20.840 --> 00:05:24.800
We kind of hear what their biggest pain points are, sketch out a few things.

131
00:05:24.840 --> 00:05:28.560
And so we've put some time and energy into it doing a little bit of breadboarding,

132
00:05:28.560 --> 00:05:30.720
a little bit of, you know, digging into their code.

133
00:05:31.120 --> 00:05:32.840
Um, sometimes it's a brand new app.

134
00:05:32.880 --> 00:05:35.240
Other times in this case, it was an existing app.

135
00:05:35.640 --> 00:05:39.000
Um, they initially came in for us to say, Hey, come and just

136
00:05:39.000 --> 00:05:40.600
rebuild our entire app, right?

137
00:05:41.000 --> 00:05:43.680
Um, that's not always the best option.

138
00:05:43.680 --> 00:05:47.640
So we initially, one of the big decisions was like, do we need to rebuild it?

139
00:05:47.640 --> 00:05:48.640
Like, is this solid?

140
00:05:48.640 --> 00:05:51.440
And we, once we got into it, we were actually like very impressed with the

141
00:05:51.440 --> 00:05:55.400
previous developers and how they had built it and the same similar tech stack to us.

142
00:05:55.400 --> 00:06:00.440
So we made a strong argument for them to keep it and just to keep it adding onto it.

143
00:06:00.440 --> 00:06:02.120
Cause it's in production.

144
00:06:02.160 --> 00:06:03.400
They've been using it for four years.

145
00:06:03.400 --> 00:06:05.120
There's a lot of value to it.

146
00:06:05.560 --> 00:06:10.440
Um, and so it's that kind of thing where just figuring out like, like kind of

147
00:06:10.440 --> 00:06:14.320
getting some strategy on like, how do they keep building like in a new,

148
00:06:14.360 --> 00:06:15.480
get them into this new process.

149
00:06:15.480 --> 00:06:18.880
We're doing a lot of education, teaching them the process of how we

150
00:06:18.960 --> 00:06:20.760
shape and frame and all those things.

151
00:06:21.200 --> 00:06:25.920
Um, but, uh, that's so that we, we kind of did this, that call, and then we kind

152
00:06:25.920 --> 00:06:29.640
of came out of that with, with like a handful of projects of which we identified

153
00:06:29.640 --> 00:06:34.840
one four week project, um, that they said, okay, this is the next most important

154
00:06:34.840 --> 00:06:35.080
thing.

155
00:06:35.080 --> 00:06:39.960
So we, we focused on that now we're like a week and a bit into that project.

156
00:06:40.560 --> 00:06:43.360
Um, which we have pretty much all of June to finish.

157
00:06:43.600 --> 00:06:45.640
Um, so we're in it right now.

158
00:06:46.080 --> 00:06:49.400
And, um, so that would be a great outcome of this call as well.

159
00:06:49.400 --> 00:06:53.800
It's just like also getting clear on that project in particular.

160
00:06:53.800 --> 00:06:58.440
So that when we leave this call, we're, we're even, you know, more clear about it.

161
00:06:59.160 --> 00:07:01.760
Um, so that'd be a by-product.

162
00:07:01.760 --> 00:07:02.440
That'd be great.

163
00:07:03.160 --> 00:07:03.960
Yeah, totally.

164
00:07:04.000 --> 00:07:09.240
Um, my first instinct in cases like this is always to go straight into the work.

165
00:07:09.600 --> 00:07:13.560
And then when we've looked, when we've all seen the same work together, then we

166
00:07:13.560 --> 00:07:19.160
can have the reflection on, could we have known something earlier if we had tried

167
00:07:19.160 --> 00:07:21.960
to do a different type of session earlier, or maybe we couldn't have even

168
00:07:21.960 --> 00:07:23.760
known that, and this is just how it is.

169
00:07:24.120 --> 00:07:25.120
Uh, you know what I mean?

170
00:07:25.120 --> 00:07:25.520
Like that?

171
00:07:25.520 --> 00:07:29.960
Cause it's, it's honestly, it's pretty hard to, for me to say something that's

172
00:07:29.960 --> 00:07:34.240
going to actually be helpful, like in, in, in general terms about process.

173
00:07:34.280 --> 00:07:34.920
Do you know what I mean?

174
00:07:35.280 --> 00:07:38.880
Um, but, um, but I'm happy to look at the diagram if you want to just kind of like.

175
00:07:39.360 --> 00:07:40.520
If you want to, you know what I mean?

176
00:07:40.520 --> 00:07:43.640
Like give it 10 minutes or whatever, and just see if there's anything to react to

177
00:07:43.640 --> 00:07:46.560
there and then I pretty much said it all.

178
00:07:46.560 --> 00:07:49.360
Um, this is just the exact thing.

179
00:07:49.360 --> 00:07:56.520
We sort of our whole pipeline here is, uh, just going through, um, Oh, sorry.

180
00:07:56.520 --> 00:07:57.000
Let's see.

181
00:07:57.560 --> 00:07:58.200
Let me leave that.

182
00:07:59.240 --> 00:07:59.440
Okay.

183
00:07:59.440 --> 00:07:59.680
Sorry.

184
00:07:59.680 --> 00:08:02.000
I have a sharing on Slack as well.

185
00:08:02.280 --> 00:08:05.280
Um, this is, you know, we do a kickoff with them.

186
00:08:05.360 --> 00:08:08.360
We do some internal sessions and then we kind of review and get two or three

187
00:08:08.360 --> 00:08:10.480
projects is basically just what I said.

188
00:08:10.720 --> 00:08:11.840
There's some details here.

189
00:08:11.880 --> 00:08:16.320
We're still kind of breadboarding this out as far as like, there's, you know,

190
00:08:16.320 --> 00:08:19.560
some, some content stuff we're handing them, like we sent them some of your

191
00:08:19.560 --> 00:08:23.800
content on YouTube, but like, you need to watch this video, like up to speed.

192
00:08:24.160 --> 00:08:28.000
Um, you know, we do this figure out lay of land problem solution, just try to like

193
00:08:28.000 --> 00:08:29.680
find out what's going to be the hard stuff.

194
00:08:30.120 --> 00:08:33.440
Um, and then going to sort of internal review.

195
00:08:33.840 --> 00:08:36.520
And so what we're finding though, is like, this kind of starts like the

196
00:08:36.520 --> 00:08:40.440
framing piece of which, you know, we still have to shape it.

197
00:08:40.440 --> 00:08:42.960
Cause there's a lot of work as we're finding working with you.

198
00:08:42.960 --> 00:08:46.120
Like this is a big, a big process, the shaping piece.

199
00:08:46.120 --> 00:08:49.520
And then there's sort of the build and the ship, which I actually made it almost

200
00:08:49.520 --> 00:08:53.440
like this, where it's like frame shape and then build ship, like that's how we teach.

201
00:08:53.480 --> 00:08:56.080
That's how, that's how we teach it in shaping in real life is that it's

202
00:08:56.080 --> 00:08:57.760
one hill and we do it the same way.

203
00:08:58.720 --> 00:08:59.200
Cool.

204
00:08:59.600 --> 00:08:59.800
Yeah.

205
00:08:59.800 --> 00:09:00.520
Very similar.

206
00:09:01.240 --> 00:09:01.560
Awesome.

207
00:09:01.560 --> 00:09:05.040
And then, and we thought like in this side of the idea, like this is where we're

208
00:09:05.240 --> 00:09:07.240
breadboarding doing design reviews.

209
00:09:07.240 --> 00:09:10.240
Like this is where we're doing super quick, you know, little prototypes

210
00:09:10.240 --> 00:09:12.480
to like get proof of concepts out.

211
00:09:13.160 --> 00:09:16.680
Um, and, and then of course this is just, you know, the, the

212
00:09:16.680 --> 00:09:18.120
build and test and ship it.

213
00:09:18.480 --> 00:09:19.000
Yeah.

214
00:09:19.360 --> 00:09:22.600
This is just, this is my thoughts of kind of how our process has changed

215
00:09:22.600 --> 00:09:26.480
because before it was shaped the entire thing, what we thought was shaping,

216
00:09:26.520 --> 00:09:27.520
which was not really shaping.

217
00:09:27.520 --> 00:09:29.320
It was like really framing.

218
00:09:29.800 --> 00:09:31.360
Um, then we would just get into it.

219
00:09:31.400 --> 00:09:34.680
We would just go from that sort of framing directly to build.

220
00:09:35.240 --> 00:09:38.720
Um, and then we would try to ship it and then we'd run into all these errors

221
00:09:38.720 --> 00:09:40.200
of like, oh no, we missed this requirement.

222
00:09:40.200 --> 00:09:41.280
We, this thing doesn't work.

223
00:09:41.280 --> 00:09:41.840
This formula.

224
00:09:41.840 --> 00:09:42.440
Exactly.

225
00:09:42.680 --> 00:09:47.720
That's the, the lack of shaping manifests as all of those time bombs going off.

226
00:09:47.760 --> 00:09:48.080
Right.

227
00:09:48.080 --> 00:09:50.040
The hidden time bombs that you didn't know were there.

228
00:09:50.040 --> 00:09:50.400
Yeah.

229
00:09:50.600 --> 00:09:50.960
Yes.

230
00:09:51.400 --> 00:09:55.600
So, um, so this was me sort of thinking of like, what's our new way of doing it?

231
00:09:55.600 --> 00:10:00.000
And this is how we've done this recent project because, um, we took them on.

232
00:10:00.000 --> 00:10:05.720
Like while but like between while we were been working with you. So we've tried to follow. This is as close as we can

233
00:10:06.760 --> 00:10:08.760
That's it. This is just my own

234
00:10:08.800 --> 00:10:15.120
Yeah, this this this in broad outlines is much closer to what to what we're teaching and shaping in real life now

235
00:10:15.920 --> 00:10:20.560
It also corresponds. I mean even when you're not working with a client when you have a

236
00:10:21.240 --> 00:10:23.960
If you're in a if you're in a in a product

237
00:10:24.640 --> 00:10:31.280
Company and you have a little bit like this like kind of like the salespeople and the rev team are kind of figuring out what they

238
00:10:31.280 --> 00:10:35.040
Want next and then you have like more the engineering side that has to go figure out how to make that happen

239
00:10:35.560 --> 00:10:40.800
What you end up having is this conversation which is very similar to the conversation you're having with the client

240
00:10:40.800 --> 00:10:46.640
Which is like what are kind of like the two or three things that are actually most important for the upcoming slice of time?

241
00:10:46.640 --> 00:10:53.080
That's going to become available right and like what's what's that thing? That's most important next and that framing step is

242
00:10:54.680 --> 00:10:56.120
It's

243
00:10:56.120 --> 00:10:57.400
mainly

244
00:10:57.400 --> 00:11:04.400
What is the what is actually most important out of the things we think we could do to do next and then the other piece?

245
00:11:04.400 --> 00:11:05.720
of it is

246
00:11:05.720 --> 00:11:10.040
There's a first it's your first line of defense of like that's insane

247
00:11:10.040 --> 00:11:13.120
That's feels like it's not even possible in a in in four weeks

248
00:11:13.120 --> 00:11:14.080
Do you know what I mean?

249
00:11:14.080 --> 00:11:17.120
Like you don't you still don't know what to do

250
00:11:17.120 --> 00:11:20.920
But it's your first line of defense about like no, we're not gonna do the entire app in four weeks

251
00:11:20.920 --> 00:11:22.240
You know what? I mean?

252
00:11:22.240 --> 00:11:27.160
And there's a little bit of a sense of like this thing that we're talking about involves

253
00:11:28.520 --> 00:11:31.000
Like some science project that we've never done before

254
00:11:31.400 --> 00:11:36.480
You know what? I mean? And then that's going to influence the way that you the way that you go into the project

255
00:11:37.440 --> 00:11:42.400
So but I think you're very much on track there and then as you have more specific questions about what about this

256
00:11:42.400 --> 00:11:44.720
What about that then we you know, then we can get into that

257
00:11:45.560 --> 00:11:47.560
Right. Okay, that's cool

258
00:11:47.680 --> 00:11:49.600
So would it be helpful to

259
00:11:49.600 --> 00:11:55.080
Just show you what we have framed and shaped so far. Yeah, exactly. Let's just look at what you have

260
00:11:55.680 --> 00:11:57.520
Okay

261
00:11:57.520 --> 00:12:00.680
So, I mean I'll share you guys this is

262
00:12:01.920 --> 00:12:05.520
Lucian's been on this this project, but I'll share real quick since I have it up

263
00:12:08.240 --> 00:12:10.240
We

264
00:12:10.440 --> 00:12:12.000
So

265
00:12:12.000 --> 00:12:13.240
generally

266
00:12:13.240 --> 00:12:16.920
We kind of just pitched like the the project here that the is that

267
00:12:17.400 --> 00:12:22.880
This is for a very company called bear tells ratings. They work with VA. So for veterans and they're helping them

268
00:12:23.400 --> 00:12:25.640
basically file a claim to to get

269
00:12:26.440 --> 00:12:27.960
compensation for

270
00:12:27.960 --> 00:12:33.040
You know and service related injuries. So the problem is that they have you know

271
00:12:33.800 --> 00:12:37.920
Potentially like 20 plus files, you know interviews with this person

272
00:12:37.920 --> 00:12:39.680
There's a lot of data around it

273
00:12:39.680 --> 00:12:42.440
and so the case managers today are like

274
00:12:42.760 --> 00:12:48.120
Pulling files down to their computer and they're like dropping them into chat GPT or and they're trying to like

275
00:12:48.400 --> 00:12:54.560
Get some value out of using AI to help summarize or just read through because some of these can be like a hundred pages long

276
00:12:55.200 --> 00:13:00.640
Uh-huh, and you can't miss like oh there was this one symptom this guy mentioned in you know

277
00:13:00.640 --> 00:13:02.920
This one call it has to be very precise

278
00:13:03.360 --> 00:13:10.240
And they're trying to basically compile all the information to then file this claim and then work with them to get it approved so

279
00:13:11.720 --> 00:13:15.560
They're they're doing like a hundred or so hundred fifty or fifty claims

280
00:13:15.560 --> 00:13:17.880
I think a month right now and they're wanting to scale that

281
00:13:18.320 --> 00:13:23.400
Right now this is where like a big bottleneck is like the case managers themselves are getting blocked

282
00:13:23.400 --> 00:13:29.280
So they identified this through a number of conversations. So once we presented this to them, they were like, yes exactly

283
00:13:29.320 --> 00:13:31.320
This is what we have

284
00:13:31.600 --> 00:13:33.600
And then

285
00:13:34.000 --> 00:13:39.320
We had some different things you're getting into different scopes, but I'll show this

286
00:13:40.360 --> 00:13:44.800
Here's like our and we do have your we got this. So we'll jump into this as well

287
00:13:45.480 --> 00:13:46.800
and

288
00:13:46.800 --> 00:13:52.120
Just kind of talk through like what's happening even today like they're uploading, you know

289
00:13:52.120 --> 00:13:57.140
We did a bit of very light breadboarding and even Lucian and I were sort of redoing all this this morning

290
00:13:57.580 --> 00:14:04.980
You know, this is this is what they do today without what you've without what you're building, uh-huh

291
00:14:04.980 --> 00:14:11.180
This is the current current. Very good. Yeah. Yeah. Yeah, and so, you know, you upload files. They upload a transcript

292
00:14:11.460 --> 00:14:13.980
they are using some open AI right now to

293
00:14:14.820 --> 00:14:19.780
create a summary or analyze or create statements and then they save that to their database and there's like a

294
00:14:20.780 --> 00:14:25.580
This is just the different tables that how everything links. I mean the in the Postgres database

295
00:14:25.900 --> 00:14:30.060
How do they save it to the database they have like some app where they have to enter stuff

296
00:14:31.220 --> 00:14:34.380
No, this is like so we started drawing it

297
00:14:36.980 --> 00:14:38.420
When the

298
00:14:38.420 --> 00:14:40.420
Transcript here's uploaded

299
00:14:40.460 --> 00:14:43.140
Yeah, let me grab this. This was supposed to be up here

300
00:14:43.140 --> 00:14:46.900
So that this is like on upload goes to s3

301
00:14:46.900 --> 00:14:52.420
They have a triggered like an event queue which then goes to processes PDFs and it does all this different

302
00:14:52.940 --> 00:14:58.300
The other things to it and then after this once it creates these sort of these outputs of these like three different

303
00:14:59.060 --> 00:15:01.060
types

304
00:15:00.000 --> 00:15:04.560
then it just saves the results basically back to the database.

305
00:15:04.560 --> 00:15:06.560
So, it's all automated.

306
00:15:06.560 --> 00:15:12.760
Uh-huh. So, what is the thing, if this starts with upload,

307
00:15:12.760 --> 00:15:16.920
so, this breadboard that we're looking at here that's with the orange pieces,

308
00:15:16.920 --> 00:15:22.520
where does this slot into the original current that you were showing me?

309
00:15:22.520 --> 00:15:24.520
I know.

310
00:15:24.520 --> 00:15:26.520
It's like this.

311
00:15:27.520 --> 00:15:31.520
So, basically, the app is here.

312
00:15:31.520 --> 00:15:35.440
They're in the interface here, and they can upload a file,

313
00:15:35.440 --> 00:15:37.520
and that would kick off this guy here.

314
00:15:37.520 --> 00:15:41.400
Right. So, the question I'm asking is, after the OpenAI step,

315
00:15:41.400 --> 00:15:49.440
when they are asking for a summary, they somehow upload that as a file?

316
00:15:49.440 --> 00:15:52.360
What's the link between the OpenAI step you showed me

317
00:15:52.360 --> 00:15:55.280
and them uploading a file into their current system?

318
00:15:55.280 --> 00:15:59.880
So, I think this is actually where OpenAI is happening,

319
00:15:59.880 --> 00:16:07.280
kind of on this processing level, because it's doing the processing here.

320
00:16:07.280 --> 00:16:08.480
There's no interaction.

321
00:16:08.480 --> 00:16:13.560
After it goes into the database, nothing else happens to it.

322
00:16:13.560 --> 00:16:15.560
That's kind of where it ends today.

323
00:16:15.560 --> 00:16:17.720
So, they, okay, hold on, I misunderstood.

324
00:16:17.720 --> 00:16:21.960
They are using OpenAI behind the scenes through the API

325
00:16:21.960 --> 00:16:24.480
in their current implementation of this.

326
00:16:24.480 --> 00:16:26.480
I see.

327
00:16:26.480 --> 00:16:30.840
There's the automatic processing of transcripts,

328
00:16:30.840 --> 00:16:38.720
and there's also a custom GPT just on the chat GPT website.

329
00:16:38.720 --> 00:16:40.720
Okay.

330
00:16:40.720 --> 00:16:44.480
So, they're uploading all this stuff.

331
00:16:44.480 --> 00:16:48.360
Then this process PDF,

332
00:16:48.360 --> 00:16:54.960
is this, are you summarizing some things that happen with user interactions,

333
00:16:54.960 --> 00:17:04.200
or is this literally all happening as an automated flow behind the scenes?

334
00:17:04.200 --> 00:17:10.920
So, for the transcripts, they are getting a transcript from their call.

335
00:17:10.920 --> 00:17:13.800
They have like a call platform, but the result of that is basically

336
00:17:13.800 --> 00:17:18.000
like a transcript comes out of that, and then that gets sent.

337
00:17:18.040 --> 00:17:26.240
So, I think it can create, it just stores summaries of that potentially in the database.

338
00:17:26.240 --> 00:17:28.800
So, yeah.

339
00:17:28.800 --> 00:17:31.400
Uh-huh, okay.

340
00:17:31.400 --> 00:17:33.720
I see that you shared a Figma with me.

341
00:17:33.720 --> 00:17:41.600
I just put that in there for anyone else who wanted to be in this instead of.

342
00:17:41.600 --> 00:17:46.160
Let me, so let's see.

343
00:17:46.160 --> 00:17:51.240
Last time when we did this, you didn't bring any breadboards in, you know,

344
00:17:51.240 --> 00:17:54.320
and so I was kind of working from a blank canvas.

345
00:17:54.320 --> 00:17:56.680
Yeah.

346
00:17:56.680 --> 00:18:02.120
What I'm inclined to do here is, how can we do this here?

347
00:18:02.120 --> 00:18:11.440
Maybe I can sign into the Figma here, and then I can maybe just kind of copy and paste,

348
00:18:11.440 --> 00:18:14.840
and we can do some things on the side.

349
00:18:14.840 --> 00:18:17.480
Yeah, you just move it to the side for sure.

350
00:18:17.480 --> 00:18:24.360
And this one, we've kind of done a few, this is like version three of this whole document.

351
00:18:24.360 --> 00:18:28.280
So, it's very flexible right now.

352
00:18:28.280 --> 00:18:34.160
Like, it's definitely not nailed down, and we sort of like pulled all the pieces apart this morning

353
00:18:34.160 --> 00:18:39.600
to take it, you know, to like reassemble this, and that's where we are at this exact moment.

354
00:18:39.600 --> 00:18:42.240
So, we jumped from that call to this call.

355
00:18:42.240 --> 00:18:43.440
Okay, cool.

356
00:18:43.440 --> 00:18:48.640
Let me just, I'm just wrestling with Figma a little bit to log in.

357
00:18:48.640 --> 00:18:51.760
It looks like I need to request access from the...

358
00:18:51.760 --> 00:18:54.880
Yeah, if you do that, here we go, approve.

359
00:18:54.880 --> 00:18:56.160
There we go.

360
00:18:56.160 --> 00:18:59.080
That's to view, let me give you...

361
00:18:59.080 --> 00:19:03.640
That's a surprisingly good workflow that actually just worked.

362
00:19:03.640 --> 00:19:04.680
That's the kind of thing that usually is.

363
00:19:04.680 --> 00:19:06.760
Can you edit though?

364
00:19:06.760 --> 00:19:09.480
Let's see.

365
00:19:09.480 --> 00:19:11.400
You can only view and comment on this board.

366
00:19:11.400 --> 00:19:13.440
Start editing.

367
00:19:13.440 --> 00:19:16.040
I'm sending a request for that.

368
00:19:16.040 --> 00:19:16.560
Oh, here we go.

369
00:19:16.560 --> 00:19:18.640
I can just set you to editor.

370
00:19:18.640 --> 00:19:19.880
Okay, it should be good now.

371
00:19:19.880 --> 00:19:24.520
It's reloading.

372
00:19:24.520 --> 00:19:25.200
Bing, bang, boom.

373
00:19:25.200 --> 00:19:27.520
Cool, okay, I see all the tools.

374
00:19:27.520 --> 00:19:30.480
I'm going to...

375
00:19:30.480 --> 00:19:31.760
I'll stop sharing.

376
00:19:31.760 --> 00:19:35.120
Okay, I'm going to take over for a moment, but then I'm going to hand it back to you

377
00:19:35.120 --> 00:19:43.280
because you're still giving me background and we're still not very far.

378
00:19:43.280 --> 00:19:48.840
But what I wanted to do was already kind of start just updating this picture

379
00:19:48.840 --> 00:19:51.880
with what's in my head.

380
00:19:51.880 --> 00:19:53.920
So this is...

381
00:19:53.920 --> 00:19:55.800
I'm just going to do a...

382
00:19:55.800 --> 00:19:58.840
So just give me a moment as I...

383
00:20:00.000 --> 00:20:06.400
I'm not completely used to to figma. This is very good. Okay, good. I'm just going to

384
00:20:08.320 --> 00:20:12.400
Can I do a is there such thing as a line? Oh, that's a connector

385
00:20:12.480 --> 00:20:18.400
Yeah, there is a line connectors, but it's hidden somewhere. Okay, i'm gonna cheat and just do this. It's perfect. Okay

386
00:20:18.960 --> 00:20:20.080
um

387
00:20:20.080 --> 00:20:21.680
and uh

388
00:20:21.680 --> 00:20:23.680
and then uh

389
00:20:24.320 --> 00:20:26.720
This is the equivalent of a frame. Yeah, this is a section

390
00:20:27.440 --> 00:20:32.640
Um, i'm just going to call this current prime. You can also do it outside of the section just to

391
00:20:33.440 --> 00:20:34.960
if you want

392
00:20:34.960 --> 00:20:36.960
Okay, and then i'm just going to grab

393
00:20:37.440 --> 00:20:39.440
This which what we was what we were looking at

394
00:20:40.320 --> 00:20:44.080
And drop it here. And i'm just trying to figure out like, um

395
00:20:45.040 --> 00:20:46.400
uh

396
00:20:46.400 --> 00:20:48.160
so

397
00:20:48.160 --> 00:20:50.160
They hit upload

398
00:20:50.560 --> 00:20:51.760
And then upload files

399
00:20:51.760 --> 00:20:57.680
This is like the dialogue where they are actually choosing the file and then they're hitting upload and then s3 upload

400
00:20:58.160 --> 00:21:04.160
Is this this is um, not actually like a user interface place. This is something that's happening behind the scenes

401
00:21:05.280 --> 00:21:06.560
behind the scenes

402
00:21:06.560 --> 00:21:08.880
Uh-huh. Good. Okay, so i'm just gonna um

403
00:21:09.760 --> 00:21:12.000
I'm, just gonna do a dumb simple thing

404
00:21:12.000 --> 00:21:15.040
I'm, just gonna make this gray and we're gonna say if it's gray

405
00:21:15.600 --> 00:21:20.320
It's not user facing and it's something that's happening behind the scenes and this is going to help us. Um,

406
00:21:20.880 --> 00:21:23.680
Uh understand a little bit better what's going on. Um,

407
00:21:24.400 --> 00:21:27.440
So, uh when you hit upload it's going to upload to

408
00:21:28.160 --> 00:21:29.760
to s3

409
00:21:29.760 --> 00:21:32.880
Um, i'm just going to write this a little bit more like as a command

410
00:21:33.520 --> 00:21:34.960
um

411
00:21:34.960 --> 00:21:36.480
and

412
00:21:36.480 --> 00:21:37.520
this

413
00:21:37.520 --> 00:21:38.640
triggered

414
00:21:38.640 --> 00:21:40.800
event queue, so, um,

415
00:21:41.440 --> 00:21:45.520
This means that now like, um, there's a there's a there's an event in the queue

416
00:21:45.600 --> 00:21:50.660
It's kind of like a job of like this this uploaded stuff hasn't been processed yet it needs to be processed

417
00:21:52.880 --> 00:22:00.640
Yeah, I see I see lucian nodding, okay good, um, so then this processing is still happening behind the scenes, yeah

418
00:22:02.640 --> 00:22:04.800
Okay, so i'm also going to change this

419
00:22:05.600 --> 00:22:06.720
um

420
00:22:06.720 --> 00:22:08.400
and um

421
00:22:08.400 --> 00:22:10.800
Uh, and this is something that we can

422
00:22:11.520 --> 00:22:13.520
unpack then so so

423
00:22:14.160 --> 00:22:16.160
This process step

424
00:22:18.320 --> 00:22:20.000
How much of this is happening

425
00:22:20.000 --> 00:22:25.280
In there's parts of this that are happening in openai and there's and then there's some output from that and then there is this

426
00:22:25.360 --> 00:22:26.880
kind of cover sheet

427
00:22:26.880 --> 00:22:29.840
so I just want to like unpack this a little bit because it kind of feels like there's a

428
00:22:30.160 --> 00:22:32.960
This to me feels murky and like there's a lot hiding in here

429
00:22:33.360 --> 00:22:36.720
And and my guess is this is where some of the complexity is. Um,

430
00:22:37.440 --> 00:22:39.440
uh, so, uh

431
00:22:40.320 --> 00:22:42.800
Creates cover sheet i'm guessing this is after

432
00:22:43.520 --> 00:22:45.520
The ai processing is done

433
00:22:48.240 --> 00:22:50.240
Yeah lucian

434
00:22:52.880 --> 00:22:57.760
The cover sheet on the uploading, uh full transparency I did like, um

435
00:22:58.560 --> 00:23:01.520
evaluate through like the transcript flow, but the

436
00:23:02.480 --> 00:23:10.720
Upload file flow that I didn't like look at that in detail. So yeah, so so should we be looking should we should we be looking?

437
00:23:11.360 --> 00:23:13.680
Yeah, sorry, should we be looking at a different flow then?

438
00:23:14.160 --> 00:23:15.040
um

439
00:23:15.040 --> 00:23:17.760
And uh, and and actually let me let me step back a second

440
00:23:17.760 --> 00:23:20.000
So the thing that I started to do here, but I want to step back

441
00:23:20.080 --> 00:23:23.440
So I don't go deep down the wrong thing. That isn't the thing that we're trying to solve

442
00:23:24.000 --> 00:23:27.120
is um, what I what I saw here was like

443
00:23:27.680 --> 00:23:30.560
Just a lack of clarity around what's front end and what's back end?

444
00:23:31.040 --> 00:23:32.320
um

445
00:23:32.320 --> 00:23:34.640
Ah, but I you did this here

446
00:23:35.280 --> 00:23:38.720
Yeah, I we started we heard you the first time you mentioned the gray

447
00:23:39.440 --> 00:23:41.280
I think we missed the

448
00:23:41.280 --> 00:23:45.680
And then I just when we started to jump around then then I just lost it, but you already did this very good

449
00:23:45.760 --> 00:23:49.520
Okay, well, no, we did we missed the s3 and the processing on the right there

450
00:23:49.600 --> 00:23:54.720
So we did it on the left, but um, I think we learned the the color coding system

451
00:23:55.280 --> 00:23:56.240
between

452
00:23:56.240 --> 00:23:58.880
The first call where we did the thing on the bottom, right? Okay

453
00:23:59.440 --> 00:24:04.560
And so this is the evolution of shaving. So so all i'm trying to do is um

454
00:24:04.880 --> 00:24:08.800
Is just get us to the place where we have like one picture where we understand what we're looking at

455
00:24:08.880 --> 00:24:09.520
Do you know what I mean?

456
00:24:09.520 --> 00:24:12.880
Because i'm I I can see that as we jump around it's a little bit like well

457
00:24:12.960 --> 00:24:17.200
There's this here and there's that there and it's just a little bit confusing. Um, uh,

458
00:24:18.000 --> 00:24:19.360
um

459
00:24:19.360 --> 00:24:22.640
Let's step back though before I continue to do that. Um, bruce

460
00:24:23.040 --> 00:24:26.100
We we started where you you said that they're they're uploading transcripts

461
00:24:26.800 --> 00:24:33.200
and then there's currently a a a back end process where some some the the

462
00:24:33.440 --> 00:24:35.680
Um open ai api is being used

463
00:24:36.960 --> 00:24:38.320
and um

464
00:24:38.320 --> 00:24:44.880
Okay, and then the place where I got stuck is that somehow something is being saved to the to the to the database and it's

465
00:24:46.960 --> 00:24:51.380
Summary analysis statements can do you guys have an understanding a little bit more concretely?

466
00:24:51.920 --> 00:24:54.800
Of what is actually being saved to the progress database?

467
00:24:54.800 --> 00:24:59.360
Do you have any sense of like what is the shape of this data or or what is this?

468
00:25:00.720 --> 00:25:02.640
That's yeah, we started. Yeah.

469
00:25:04.320 --> 00:25:09.200
Um, the summary of the transcript, the transcripts are like totally different from like

470
00:25:09.200 --> 00:25:13.440
files. Um, it's just like, they had a meeting, you know, like through a hundred and less

471
00:25:14.160 --> 00:25:20.320
and, uh, summary and they extract like claims lists. So basically you can make like a claim

472
00:25:20.320 --> 00:25:25.920
for like knee injury and PTSD and whatever. Uh, and then statements is like, kind of like drafting,

473
00:25:25.920 --> 00:25:30.160
like, Oh, I did a service. So like in the first person, like the vet is saying, like,

474
00:25:30.160 --> 00:25:35.920
I was in Afghanistan, like there was an explosion. Now I'm afraid and blah, blah, blah. So kind of

475
00:25:35.920 --> 00:25:41.120
like all the artifacts that come out of the transcript summarization. I see this.

476
00:25:41.120 --> 00:25:46.960
We started sketching out the database here, which we did in another project. And as we just started

477
00:25:46.960 --> 00:25:52.000
doing that now, but you can see like, there's the claims add ons, there's the meeting transcript.

478
00:25:52.400 --> 00:25:56.640
Um, I see. Okay. Where, where's that sketch out of the database?

479
00:25:56.640 --> 00:26:01.440
Uh, what you just pasted, uh, over there on the right. Yeah. Down here. So you see the,

480
00:26:01.440 --> 00:26:07.440
in the gray, like claim add ons is a, is a table claims is a table. Um, meeting transcript is a

481
00:26:07.440 --> 00:26:10.960
table and you can just, we did some of the links so we could see like a claim ID links to claim

482
00:26:11.600 --> 00:26:17.520
ID, you know, uh, it's not as complete as I would like it, but there's like the general idea. Like

483
00:26:17.520 --> 00:26:24.640
there's like multiple places that I see. Yeah. I see. So, um, okay. So I'm just going to order

484
00:26:24.640 --> 00:26:30.240
this a little bit. Um, and, uh, now this, this is your documentation of how it works currently.

485
00:26:31.120 --> 00:26:38.160
Yep. Yes. Yeah. Got it. Okay. Um, can I do a section in a section? Let's try. Um,

486
00:26:38.160 --> 00:26:52.400
um, um, so these are claim at, there are, there are claims with claim add ons and there are

487
00:26:52.400 --> 00:26:59.040
transcripts and there are users. How do these things, these things, do they map to what you're

488
00:26:59.040 --> 00:27:03.840
saying as summary analysis and statements, or is there still more work to, to, to spell out

489
00:27:04.400 --> 00:27:07.840
what the relationship is between this and summary analysis and statements?

490
00:27:10.160 --> 00:27:18.480
Um, like summary goes on the, like meeting transcripts, uh, table. Uh huh. Okay. And also

491
00:27:18.480 --> 00:27:26.960
on the claim add ons table, uh, analysis goes on like claim add ons and statements goes on claim

492
00:27:27.040 --> 00:27:35.360
add ons. Uh huh. Okay. Uh huh. So, um, my instinct here is, um,

493
00:27:38.240 --> 00:27:43.440
what I want to do is I'm just going to, I'm just going to change Postgres DB to, I'm just going

494
00:27:43.440 --> 00:27:51.280
to say like, what needs to happen is like save to Postgres DB. And I'm just going to describe,

495
00:27:51.840 --> 00:28:03.280
um, summary and, um, uh, okay. Summary and, um, claim claims list,

496
00:28:04.240 --> 00:28:09.920
summary claims list and statements as claim add ons.

497
00:28:12.480 --> 00:28:19.280
Yep. Wow. So many interesting UI elements on this Figma. Okay. Summary, summary claims list

498
00:28:19.280 --> 00:28:28.640
and statements as, as claim add ons and, um, transcripts. That was another thing here. Um,

499
00:28:29.200 --> 00:28:36.560
uh, summary analysis and statements and the transcripts, which we uploaded those anyway,

500
00:28:36.560 --> 00:28:42.560
actually before going to the, were they anyway already? Um, um, but we just had this earlier.

501
00:28:43.280 --> 00:28:46.640
Yeah. We just had the summary to the transcript that already exists.

502
00:28:50.240 --> 00:28:53.920
Uh, uh huh. Okay. So we're going to, so, so we're actually saving the,

503
00:28:54.560 --> 00:28:58.320
and I'm, I'm, I'm just being super informal here. You know what I mean? Like I'm saying like,

504
00:28:59.040 --> 00:29:05.840
um, uh, this there's some work summarize transcripts. So this, is this a, is this a,

505
00:29:05.840 --> 00:29:10.880
um, what is this meant to be saying? Is this a call to summarize? And these are arguments to

506
00:29:10.880 --> 00:29:15.120
that? Or is this a list of things that are going to kind of happen over here?

507
00:29:15.280 --> 00:29:22.080
Yeah. There's like a event queue, you know? Um, so it's like summarizing the transcript. And then

508
00:29:22.080 --> 00:29:27.440
if it can like, um, I'm probably not saying like super accurate, what it may be the process

509
00:29:27.440 --> 00:29:33.840
slightly different, but it's like, uh, first summarizing. And then there's like a thing

510
00:29:33.840 --> 00:29:40.480
called analyze, which is like, get the claims list. Uh, and then for each claim, like knee,

511
00:29:40.480 --> 00:29:48.880
PTSD, et cetera, it makes, uh, a summary of like the part of the conversation that was talking

512
00:29:48.880 --> 00:29:58.960
about that particular issue and then make statement for each like claim. I see. I see. Okay. Um,

513
00:30:00.000 --> 00:30:06.600
And these are each lined up as jobs here and then performed later, or is this

514
00:30:06.680 --> 00:30:12.600
all happening just kind of as one pipeline where one thing triggers off the next thing?

515
00:30:15.520 --> 00:30:18.120
Like the pipeline.

516
00:30:18.840 --> 00:30:19.280
Uh-huh.

517
00:30:19.320 --> 00:30:19.520
Good.

518
00:30:19.560 --> 00:30:19.920
Okay.

519
00:30:20.200 --> 00:30:29.760
So the way that I would be trying to clarify this in my head right now,

520
00:30:29.760 --> 00:30:37.160
or what I'm trying to do to clarify this in my head right now is, I want to sort

521
00:30:37.160 --> 00:30:42.080
of see all the backend pieces broken out the same way that we break out the front

522
00:30:42.080 --> 00:30:42.600
end pieces.

523
00:30:42.600 --> 00:30:45.440
You know, when we see the front end pieces, it's really like ding, ding, ding.

524
00:30:45.440 --> 00:30:50.840
Like we, we see how all the, so what I want to do is instead of kind of

525
00:30:50.840 --> 00:30:53.920
summarizing it and then it's all like in our head, you know what I mean?

526
00:30:54.240 --> 00:30:57.040
I want to think that there's, there's basically, there's a, there's going to

527
00:30:57.040 --> 00:31:00.440
be like a summarized transcript and I'm just going to annotate it like this to say

528
00:31:00.440 --> 00:31:02.120
that this is a function that's going to happen.

529
00:31:03.080 --> 00:31:11.320
There's going to be an analysis of transcript to that, that produces the

530
00:31:11.360 --> 00:31:12.280
claims list.

531
00:31:12.360 --> 00:31:12.920
Yeah.

532
00:31:13.000 --> 00:31:23.320
As the, so I'm going to say like analysis, like analyze transcript is going to like

533
00:31:23.320 --> 00:31:25.600
output like a set of claims.

534
00:31:26.120 --> 00:31:33.040
This, I don't actually know what's coming out of this is the summary, like a single

535
00:31:33.040 --> 00:31:33.960
block of text.

536
00:31:36.040 --> 00:31:36.360
Yeah.

537
00:31:36.360 --> 00:31:36.680
Yeah.

538
00:31:36.720 --> 00:31:36.960
Yeah.

539
00:31:37.000 --> 00:31:42.040
It just like kind of removes like arms, butts and like, how were you doing this

540
00:31:42.040 --> 00:31:42.600
morning?

541
00:31:42.640 --> 00:31:47.080
Like, ah, so it's basically like transcript in transcript out, but cleaned up.

542
00:31:48.200 --> 00:31:48.800
Yeah.

543
00:31:49.800 --> 00:31:50.480
Yeah.

544
00:31:55.720 --> 00:31:56.200
Okay.

545
00:31:57.000 --> 00:32:05.760
Um, uh, and then what we have is, um, so we should find, we should feel that we

546
00:32:05.760 --> 00:32:08.160
don't need this and we don't need this when we start to come to the end of this.

547
00:32:08.160 --> 00:32:08.520
Right.

548
00:32:08.840 --> 00:32:12.400
We should start to feel like this is happening first.

549
00:32:12.920 --> 00:32:17.920
Uh, then after we've cleaned the transcript, then we're getting claims out

550
00:32:17.920 --> 00:32:19.280
of it, right?

551
00:32:19.440 --> 00:32:22.480
So we're actually kind of taking the cleaned transcript as the input to the

552
00:32:22.480 --> 00:32:23.440
analyze step.

553
00:32:25.120 --> 00:32:25.640
Yep.

554
00:32:26.080 --> 00:32:39.840
And then from there, then, um, for each claim, summarize the claim and, um, uh,

555
00:32:40.440 --> 00:32:47.240
and the attachment, the, the summary is, is added to the claim as an add on.

556
00:32:47.400 --> 00:32:47.720
Yeah.

557
00:32:47.760 --> 00:32:48.800
As a claim add on.

558
00:32:50.240 --> 00:32:50.520
Yep.

559
00:32:53.000 --> 00:32:53.280
Yeah.

560
00:32:53.280 --> 00:32:55.640
There's the summary and there's the statement.

561
00:32:56.200 --> 00:32:56.440
Mm.

562
00:32:56.440 --> 00:32:56.760
Hmm.

563
00:32:56.840 --> 00:32:57.080
Mm.

564
00:32:57.080 --> 00:32:57.320
Hmm.

565
00:32:59.960 --> 00:33:06.200
Um, again, it's like a little bit messy, but what I'm trying to get to the thing

566
00:33:06.800 --> 00:33:11.760
that's important here that I'm trying to see is, um, is the inputs and the outputs

567
00:33:11.760 --> 00:33:13.880
and what the chunks of work are, right?

568
00:33:13.880 --> 00:33:16.920
Cause this is going to help us to understand like where we intervene and if

569
00:33:16.920 --> 00:33:18.400
we're doing all the same things or not.

570
00:33:18.440 --> 00:33:21.520
And, uh, um, and we're almost to the end of this.

571
00:33:21.520 --> 00:33:24.160
And then I think we can lift up our heads and see where you guys are trying to get

572
00:33:24.160 --> 00:33:24.680
to next.

573
00:33:24.880 --> 00:33:30.400
So there's, um, so then also for each claim, I'm also, I'm pulling these out as

574
00:33:30.400 --> 00:33:33.960
two different things, just cause they're two different processing steps is I want

575
00:33:33.960 --> 00:33:39.360
to, um, um, uh, well, yeah.

576
00:33:39.360 --> 00:33:40.960
So I'm generating statements.

577
00:33:41.000 --> 00:33:41.760
This is also funny.

578
00:33:41.760 --> 00:33:43.720
So like I'm summarizing the claim.

579
00:33:43.720 --> 00:33:48.880
I'm actually taking the transcript in with the claim in order to do this.

580
00:33:50.240 --> 00:33:56.640
By the way, pretty much just the transcript and well, the claim too, is

581
00:33:56.640 --> 00:33:59.520
like, you know, we're talking to just about me because we have to know which

582
00:33:59.520 --> 00:34:03.600
claim we're talking about and then, and then in this transcript, then we're

583
00:34:03.600 --> 00:34:08.080
going to come out of the book with a claim add-on, um, we're going to, we're

584
00:34:08.080 --> 00:34:12.840
going to, we're going to add a, um, summary as claim add-on it's like what

585
00:34:12.840 --> 00:34:13.639
comes out of this.

586
00:34:13.960 --> 00:34:14.460
Yeah.

587
00:34:15.239 --> 00:34:15.739
Yeah.

588
00:34:16.880 --> 00:34:19.280
And, uh, and then the last piece here,

589
00:34:25.560 --> 00:34:29.159
I think we can actually, uh,

590
00:34:29.560 --> 00:34:33.719
start to start to, then the last thing we need to do is generate statements.

591
00:34:33.719 --> 00:34:40.199
And so, so generating statements means, um, that we are adding claim add-ons

592
00:34:40.199 --> 00:34:45.639
again, um, which were statements that had to do with that claim from the transcript.

593
00:34:47.239 --> 00:34:51.199
Uh, they basically take the same input, like the claim itself and the transcript.

594
00:34:51.960 --> 00:34:52.460
Mm.

595
00:34:53.880 --> 00:34:55.639
As far as I remember, it's like that.

596
00:34:55.639 --> 00:34:56.139
Yeah.

597
00:34:56.139 --> 00:34:56.639
Mm hmm.

598
00:35:00.000 --> 00:35:04.160
And again, it's outputting an add-on to the claim?

599
00:35:04.160 --> 00:35:05.000
Yeah.

600
00:35:08.260 --> 00:35:09.100
Okay.

601
00:35:10.620 --> 00:35:12.080
And at each step,

602
00:35:13.480 --> 00:35:15.940
whenever we have claim add-ons are getting created,

603
00:35:15.940 --> 00:35:19.520
I mean, that's like the Postgres database

604
00:35:19.520 --> 00:35:21.840
is actually getting written to all the time

605
00:35:21.840 --> 00:35:23.560
as these things happen, right?

606
00:35:23.560 --> 00:35:25.000
It's not like all this has to happen

607
00:35:25.000 --> 00:35:27.880
and then we have to do it as a batch, right?

608
00:35:27.880 --> 00:35:28.720
Yeah.

609
00:35:28.720 --> 00:35:29.540
Okay.

610
00:35:30.440 --> 00:35:31.280
So,

611
00:35:35.080 --> 00:35:37.040
okay, so now

612
00:35:39.240 --> 00:35:40.080
we have,

613
00:35:44.700 --> 00:35:46.280
I think what we're gonna find is this becomes

614
00:35:46.280 --> 00:35:47.680
a little bit easier to make sense out of

615
00:35:47.680 --> 00:35:51.340
because now we have a workflow that's happening here.

616
00:35:53.600 --> 00:35:55.240
What happens when this is done?

617
00:35:56.380 --> 00:35:57.220
Let's see.

618
00:35:57.980 --> 00:36:00.780
We didn't have a definition of what was done then.

619
00:36:00.780 --> 00:36:04.220
So, there reaches a point where like all of this happens

620
00:36:04.220 --> 00:36:07.020
and then if we summarize this,

621
00:36:07.020 --> 00:36:09.260
so if I just wrap this in something

622
00:36:10.260 --> 00:36:12.720
and I'm just gonna say all of this is like,

623
00:36:19.860 --> 00:36:21.460
what did you guys call this?

624
00:36:21.460 --> 00:36:23.200
Process transcript, maybe?

625
00:36:23.200 --> 00:36:24.360
Yeah, process,

626
00:36:25.200 --> 00:36:27.860
process transcript.

627
00:36:27.860 --> 00:36:29.840
Like this whole thing

628
00:36:31.320 --> 00:36:32.560
takes in,

629
00:36:33.920 --> 00:36:35.600
it takes in a transcript

630
00:36:40.240 --> 00:36:42.700
and what do we get out, right?

631
00:36:42.700 --> 00:36:46.760
Like what we get out is we have

632
00:36:48.540 --> 00:36:51.840
claim with add-ons for

633
00:36:54.760 --> 00:36:58.760
summaries and statements, let's say.

634
00:37:02.200 --> 00:37:03.040
Yep.

635
00:37:04.620 --> 00:37:09.080
So, this is actually,

636
00:37:11.040 --> 00:37:15.680
so before we had transcripts tab was kind of pointing to

637
00:37:20.200 --> 00:37:21.460
the database.

638
00:37:21.460 --> 00:37:23.520
It queries the database.

639
00:37:24.000 --> 00:37:24.840
Okay.

640
00:37:24.840 --> 00:37:27.600
So, a different way of saying that is that,

641
00:37:27.600 --> 00:37:30.680
so this is where we're gonna read what was done here.

642
00:37:30.680 --> 00:37:31.520
Is that right?

643
00:37:32.960 --> 00:37:34.960
As far as I know, only the summaries,

644
00:37:34.960 --> 00:37:36.600
like the main transcript summary

645
00:37:36.600 --> 00:37:39.960
and the like per claim transcript summaries,

646
00:37:39.960 --> 00:37:41.800
but not the statements.

647
00:37:42.720 --> 00:37:44.120
That's kind of what I did.

648
00:37:45.040 --> 00:37:46.480
Uh-huh, okay.

649
00:37:46.480 --> 00:37:51.480
So, the place where I can end up here

650
00:37:51.800 --> 00:37:55.080
and then we can take a pause is that

651
00:37:55.080 --> 00:37:57.020
actually like all of this,

652
00:37:58.300 --> 00:38:01.080
the shape of this data that like kind of comes out of this,

653
00:38:01.080 --> 00:38:06.080
like matters as an input to this like summaries here, right?

654
00:38:06.160 --> 00:38:09.560
This is like when you're trying to describe

655
00:38:09.560 --> 00:38:11.760
kind of like how,

656
00:38:13.400 --> 00:38:16.040
what data is shown in the summaries, right?

657
00:38:17.000 --> 00:38:17.840
Sure.

658
00:38:17.840 --> 00:38:18.660
So,

659
00:38:19.660 --> 00:38:24.700
so let's just pause there

660
00:38:26.220 --> 00:38:28.620
and say that all of this, I don't think I can,

661
00:38:29.620 --> 00:38:30.740
is it gonna let me do,

662
00:38:30.740 --> 00:38:33.120
oh yeah, we can do something like this.

663
00:38:33.120 --> 00:38:34.880
So, more or less, we're gonna get this.

664
00:38:34.880 --> 00:38:39.180
So, here, I'm just doing this very clunky trick of like,

665
00:38:39.180 --> 00:38:41.100
if it's input, I'm just kind of upper lefting it

666
00:38:41.100 --> 00:38:43.220
above something, do you know what I mean?

667
00:38:43.220 --> 00:38:45.960
And I'm just thinking of this as a read model.

668
00:38:45.960 --> 00:38:47.720
I mean, if we really wanted to color code things,

669
00:38:47.720 --> 00:38:48.560
we can do that more,

670
00:38:48.560 --> 00:38:50.120
but I don't wanna waste time with diagramming.

671
00:38:50.120 --> 00:38:52.360
I wanna kind of like keep getting to the actual problems

672
00:38:52.360 --> 00:38:53.180
that you guys have.

673
00:38:53.180 --> 00:38:56.960
So, let's say that,

674
00:38:56.960 --> 00:39:00.200
so now what we've got is we've got a clearer sense

675
00:39:00.200 --> 00:39:02.500
of what's happening with the data and the processing steps

676
00:39:02.500 --> 00:39:03.340
and where it takes us.

677
00:39:03.340 --> 00:39:05.360
And then we end up being able to look at summaries

678
00:39:05.360 --> 00:39:06.860
on the transcripts tab,

679
00:39:06.860 --> 00:39:11.860
because we have the summaries as part of the claim data.

680
00:39:15.520 --> 00:39:17.600
Is there anything left here that is a mess?

681
00:39:18.440 --> 00:39:19.280
No, transcripts have summaries.

682
00:39:19.280 --> 00:39:20.120
Okay, good.

683
00:39:20.120 --> 00:39:22.880
So, I'm just gonna delete this, delete that.

684
00:39:22.880 --> 00:39:26.180
And then the last piece is there's an edit from summaries.

685
00:39:28.160 --> 00:39:29.000
Yeah.

686
00:39:29.000 --> 00:39:33.520
And before we keep digging,

687
00:39:33.520 --> 00:39:36.600
let's just say that we're just back up to date,

688
00:39:36.600 --> 00:39:38.240
Bruce, with what you were trying to walk us through

689
00:39:38.240 --> 00:39:39.080
in the beginning.

690
00:39:40.440 --> 00:39:43.000
Okay, so we took a little time just to get clearer.

691
00:39:44.880 --> 00:39:46.080
So, what are you trying to do?

692
00:39:46.080 --> 00:39:47.780
We're still in the current world.

693
00:39:50.280 --> 00:39:51.120
Nice.

694
00:39:51.120 --> 00:39:53.000
This is current, yeah, current.

695
00:39:53.000 --> 00:39:55.840
And I think the piece that,

696
00:39:55.840 --> 00:39:58.160
so there's this whole transcript piece,

697
00:39:58.160 --> 00:40:00.080
which is like one half of this.

698
00:40:00.000 --> 00:40:04.260
then there's the files, um, uploading, which is still, um, important.

699
00:40:04.260 --> 00:40:07.580
Cause we, um, there's a whole process that happens on the transcript.

700
00:40:07.580 --> 00:40:10.380
And then I started drawing it out to the right of what you, where you're

701
00:40:10.380 --> 00:40:11.840
working with the color code.

702
00:40:11.840 --> 00:40:15.240
I think if you go to the, uh, to the right, uh, this is the upload files here.

703
00:40:15.580 --> 00:40:15.780
Yeah.

704
00:40:15.780 --> 00:40:19.020
But if you go to the right of where you were, I'd kind of dragged it over and

705
00:40:19.020 --> 00:40:21.240
started fixing the colors here.

706
00:40:21.360 --> 00:40:24.680
And so there's these like logic things here.

707
00:40:24.680 --> 00:40:28.760
Like if PDF do this, if that, and they have these little functions

708
00:40:28.760 --> 00:40:36.220
called like handle evidence, PDF handle doc, and then, um, so these

709
00:40:36.220 --> 00:40:40.300
pieces are like basically based off some sort of filter logic, like

710
00:40:40.300 --> 00:40:43.160
depending on what type of file is there, it'll do different things to it.

711
00:40:43.540 --> 00:40:43.760
Yeah.

712
00:40:43.840 --> 00:40:46.880
Now, of course we would go through and I need to figure out like what we just

713
00:40:46.880 --> 00:40:49.200
did now, like what does the handle PDF do?

714
00:40:49.200 --> 00:40:52.560
Oh, well, it, it looks for this, it does that it does whatever.

715
00:40:52.880 --> 00:40:58.640
Um, uh, and, and I'm going to just, I'm grabbing it.

716
00:40:58.640 --> 00:41:04.940
Like, so we're, we're, um, I'm pulling this out of like, uh, the code base now.

717
00:41:04.940 --> 00:41:05.960
And this, so what, okay.

718
00:41:05.960 --> 00:41:10.320
So if we kind of pause and kind of do our, uh, step back and talk

719
00:41:10.320 --> 00:41:11.600
about the game for a second.

720
00:41:11.760 --> 00:41:12.040
Yeah.

721
00:41:12.040 --> 00:41:12.480
Game off.

722
00:41:12.480 --> 00:41:12.800
Okay.

723
00:41:13.160 --> 00:41:13.560
Game off.

724
00:41:13.560 --> 00:41:17.920
Like it's, uh, so, so the unique thing here is that we are

725
00:41:17.920 --> 00:41:19.360
working on someone else's app.

726
00:41:19.640 --> 00:41:23.040
Like we always build our own stuff before this.

727
00:41:23.160 --> 00:41:27.640
So one very big rabbit hole here is we've just never worked on this app.

728
00:41:27.640 --> 00:41:29.560
And so on face value, you're like, cool.

729
00:41:29.560 --> 00:41:30.480
It's the same tech stack.

730
00:41:30.480 --> 00:41:31.160
It's JavaScript.

731
00:41:31.160 --> 00:41:31.840
This makes sense.

732
00:41:31.880 --> 00:41:33.520
It's using all the same tools we use.

733
00:41:33.680 --> 00:41:34.080
Got it.

734
00:41:36.560 --> 00:41:38.680
It's only the, only the beginning of the mystery.

735
00:41:39.880 --> 00:41:44.520
And, uh, you know, it seems, it seems clean and organized and it's well.

736
00:41:44.760 --> 00:41:47.360
Um, structured, like it's not, it doesn't look sloppy.

737
00:41:47.360 --> 00:41:52.160
It doesn't look like, you know, uh, the output, like an AI generated app.

738
00:41:52.200 --> 00:41:53.520
Um, so there's someone did it.

739
00:41:53.520 --> 00:41:54.320
He did it by hand.

740
00:41:54.400 --> 00:41:55.600
They've been tweaking it.

741
00:41:55.600 --> 00:41:59.880
And to me, the, one of the most valuable things about an app is the longevity.

742
00:41:59.880 --> 00:42:05.160
So if it's been in use in production in the business for four years, like that,

743
00:42:05.240 --> 00:42:08.920
that means something just that they've worked out a lot of the quirky bits.

744
00:42:09.000 --> 00:42:10.000
Um, it may not be right.

745
00:42:10.000 --> 00:42:13.640
Like the workflow may be wrong, but it's got some, it's

746
00:42:13.640 --> 00:42:15.200
got some, uh, some years on it.

747
00:42:15.200 --> 00:42:20.760
So, um, so, so that's the part now we're trying to like pick apart.

748
00:42:21.520 --> 00:42:26.160
What was the, what's the logic of how it currently works is clearly like, um,

749
00:42:26.160 --> 00:42:29.360
and I'm not trying to like, I'm not panicking, I'm not panicking at all.

750
00:42:29.360 --> 00:42:32.120
If anyone's wondering, I'm not panicking because we are two weeks into this

751
00:42:32.120 --> 00:42:37.920
project that's already paid for, and we are, um, uh, trying to figure

752
00:42:37.920 --> 00:42:39.720
out what all these functions are.

753
00:42:39.720 --> 00:42:43.640
So we, we, one massive thing I'm seeing we missed is we just like, did not go

754
00:42:43.640 --> 00:42:48.000
deep enough into the existing code to really document, like, cause we haven't

755
00:42:48.000 --> 00:42:49.520
even talked about the project yet.

756
00:42:49.520 --> 00:42:52.800
We're just trying to show you right now, like what the current version is.

757
00:42:52.800 --> 00:42:56.400
And we're still struggling to tell you what, like what pieces are.

758
00:42:56.400 --> 00:43:00.400
So I've been actually the least involved in this, uh, between Chris and, uh,

759
00:43:00.440 --> 00:43:02.640
Lucian, there's another developer, MJ, who's on it.

760
00:43:03.160 --> 00:43:08.040
Um, so, uh, we, we do need to go, like we need, that's one thing in our process

761
00:43:08.040 --> 00:43:12.120
I'm seeing is that we just didn't go that deep into the code to figure out

762
00:43:12.600 --> 00:43:16.360
we should spend as much time documenting the current version as we

763
00:43:16.360 --> 00:43:17.800
should almost documenting the new.

764
00:43:18.200 --> 00:43:19.200
And we kind of like,

765
00:43:19.480 --> 00:43:25.000
Yeah, the new will be, the new will feel very easy to describe when the

766
00:43:25.000 --> 00:43:29.600
current is well-defined because the new is just going to be this little diff of

767
00:43:29.600 --> 00:43:33.080
like, oh, we, so what we're going to get to is we're going to get to like, oh,

768
00:43:33.080 --> 00:43:36.480
we're going to, we're going to slice here and we're going to slice here.

769
00:43:36.480 --> 00:43:38.240
And then this thing is going to be different.

770
00:43:38.240 --> 00:43:38.880
You know what I mean?

771
00:43:38.920 --> 00:43:43.480
And you're going to be kind of like describing the interventions, um, and

772
00:43:43.480 --> 00:43:47.840
you're going to have far fewer surprises and, and, uh, you're going to know how

773
00:43:47.840 --> 00:43:51.760
to react when the, uh, the things that you inevitably missed come up because

774
00:43:51.760 --> 00:43:53.000
you can see where they belong.

775
00:43:54.080 --> 00:43:54.440
Right.

776
00:43:55.280 --> 00:43:57.240
And I, that's exactly.

777
00:43:57.240 --> 00:44:01.800
So I think, um, that, that was, that was the exact perfect.

778
00:44:01.800 --> 00:44:03.280
What we needed right there was yes.

779
00:44:03.280 --> 00:44:08.560
It's the, it's really like if we can spend 80, 70, 80% of the time on just

780
00:44:08.560 --> 00:44:12.240
getting super clear about the current, because I think when we had to explain

781
00:44:12.240 --> 00:44:16.840
the DM example to you, like to us, we just knew what DMS were and we just were

782
00:44:16.840 --> 00:44:18.920
like, no, we're going to add this button and we're going to do this thing.

783
00:44:18.920 --> 00:44:20.040
And then we were like, whoa, whoa, whoa.

784
00:44:20.040 --> 00:44:21.360
Like how does it work right now?

785
00:44:21.360 --> 00:44:23.680
And we had to go through, we had to actually document and that software,

786
00:44:23.680 --> 00:44:27.160
like we built ourselves and we hadn't documented the current.

787
00:44:27.160 --> 00:44:31.080
So it's not, I wouldn't even say it's necessarily like our appers and someone

788
00:44:31.080 --> 00:44:31.760
else's app.

789
00:44:31.880 --> 00:44:37.240
Um, but I think that's very helpful mindset to say, like spend a lot of time

790
00:44:37.240 --> 00:44:39.200
documenting how it is currently.

791
00:44:39.600 --> 00:44:44.080
Um, and Ira and I were working on a totally different app, different clients.

792
00:44:44.160 --> 00:44:46.880
Um, and we actually spent, and he was new to the whole project, which actually

793
00:44:46.880 --> 00:44:52.080
was very helpful for us because we ended up having a lot of sort of dumb questions.

794
00:44:52.200 --> 00:44:54.080
Uh, you know, the best questions.

795
00:44:54.160 --> 00:44:54.480
Yeah.

796
00:44:54.800 --> 00:44:55.040
Yeah.

797
00:44:55.040 --> 00:44:58.880
So he was like, so what's a product, like what's a variant, like how does it relate?

798
00:44:58.880 --> 00:45:00.080
And we, and then all of a sudden.

799
00:45:00.000 --> 00:45:01.960
Even I've been working on it for six months.

800
00:45:01.960 --> 00:45:02.920
I was like, Oh yeah.

801
00:45:02.920 --> 00:45:05.080
So it's like a, you know, it looks like this.

802
00:45:05.080 --> 00:45:08.760
And we had these little sketches on the, on the white board to, to make it easy.

803
00:45:08.760 --> 00:45:12.280
And so we all got really clear cause we, we had to document, you

804
00:45:12.280 --> 00:45:13.840
know, how the database was today.

805
00:45:13.840 --> 00:45:15.120
Like what, and then what's that?

806
00:45:15.120 --> 00:45:15.300
Okay.

807
00:45:15.300 --> 00:45:17.160
We're going to add a new table here and we're going to

808
00:45:17.160 --> 00:45:18.480
connect this table to that.

809
00:45:18.480 --> 00:45:21.280
It became very clear what the diff is going to be.

810
00:45:21.640 --> 00:45:23.120
And that was very helpful.

811
00:45:23.240 --> 00:45:26.640
Um, that was, that was worth the first 45 minutes.

812
00:45:28.080 --> 00:45:28.520
It was great.

813
00:45:28.560 --> 00:45:29.040
Yeah.

814
00:45:29.080 --> 00:45:29.320
Yeah.

815
00:45:29.320 --> 00:45:29.880
That's good.

816
00:45:31.000 --> 00:45:37.400
Um, so, so, uh, one more thing on the game off here is, um, um, you'll find

817
00:45:37.400 --> 00:45:41.240
that, you know, like if we felt like we had to document the entire app, we

818
00:45:41.240 --> 00:45:42.440
would all want to kill ourselves.

819
00:45:42.720 --> 00:45:47.240
But like, if we, when we know that this is the part of app I'm supposed to be

820
00:45:47.240 --> 00:45:51.800
working on, then documenting, it should feel like it shouldn't feel like

821
00:45:51.800 --> 00:45:53.680
this annoying, tedious thing.

822
00:45:53.680 --> 00:45:58.280
It should feel like, Oh, like I'm untangling the knot, you know?

823
00:45:58.360 --> 00:46:00.320
And like, this is my knot.

824
00:46:00.360 --> 00:46:01.960
I'm like supposed to be working in here.

825
00:46:01.960 --> 00:46:04.120
So like, I'm getting, I'm getting clearer.

826
00:46:04.120 --> 00:46:05.040
I'm getting closer.

827
00:46:05.040 --> 00:46:06.080
This feels like progress.

828
00:46:06.120 --> 00:46:06.440
Right.

829
00:46:07.560 --> 00:46:08.120
Um,

830
00:46:09.520 --> 00:46:10.520
a follow up question.

831
00:46:10.600 --> 00:46:10.800
Yeah.

832
00:46:12.240 --> 00:46:12.520
Yeah.

833
00:46:12.800 --> 00:46:16.240
Uh, so speaking of like untangling the knot, it's kind of what you do with

834
00:46:16.240 --> 00:46:19.560
the process transcript, because all, it seemed like a summary of that

835
00:46:19.560 --> 00:46:22.120
information was in the, in the arrow.

836
00:46:22.360 --> 00:46:22.800
Right.

837
00:46:23.280 --> 00:46:23.920
Exactly.

838
00:46:23.960 --> 00:46:24.240
Yeah.

839
00:46:24.280 --> 00:46:24.760
Yeah.

840
00:46:24.920 --> 00:46:28.360
So I, I, my question is, I'm just kind of coming from that.

841
00:46:28.680 --> 00:46:34.840
Um, how do you think about like arrows and putting text over them?

842
00:46:35.400 --> 00:46:39.080
And is it appropriate or when should it actually be unraveled?

843
00:46:39.760 --> 00:46:44.920
So first of all, I have to tell you that, um, the, uh, this

844
00:46:44.920 --> 00:46:48.480
breadboarding method started with just defining the UI.

845
00:46:48.960 --> 00:46:52.560
And when we were just using it, when we just use it for UI,

846
00:46:55.760 --> 00:46:57.280
we would, might do something like this.

847
00:46:57.440 --> 00:47:02.920
Um, um, uh, I'm just going to make up something.

848
00:47:02.920 --> 00:47:05.800
Let's say there's some other, let's say there's an upload file button.

849
00:47:06.280 --> 00:47:08.840
And then I select the file and then I have like an upload button

850
00:47:08.840 --> 00:47:10.120
from the modal here, you know?

851
00:47:10.440 --> 00:47:13.760
And then like, let's say like, after I upload, I'm going to land on the

852
00:47:13.760 --> 00:47:16.120
index of files and I'm going to see all the files I uploaded.

853
00:47:16.120 --> 00:47:17.560
Let's say it's as simple as that.

854
00:47:18.480 --> 00:47:19.000
Do you know what I mean?

855
00:47:19.480 --> 00:47:26.920
Um, I could say here, um, store the files route to index.

856
00:47:28.480 --> 00:47:32.720
And when I do this, what I'm saying is like, um, I'm only

857
00:47:32.720 --> 00:47:34.080
showing the interface elements.

858
00:47:34.080 --> 00:47:37.560
I like to think that like anything that's in a box is something that

859
00:47:37.560 --> 00:47:39.200
I'm like actually spelling out.

860
00:47:40.160 --> 00:47:43.600
And, and this is like, like a little shortcut of saying, well, like, and

861
00:47:43.600 --> 00:47:47.440
along the way, this is what happens behind the scenes and there's good.

862
00:47:47.440 --> 00:47:50.120
There's always going to be things that aren't worth spelling out because

863
00:47:50.120 --> 00:47:53.040
you've already seen it a million times before and there's nothing interesting

864
00:47:53.040 --> 00:47:54.080
there, do you know what I mean?

865
00:47:54.360 --> 00:47:57.720
And so like, here's a case where I wouldn't stress about that at all.

866
00:47:57.960 --> 00:48:05.440
But when I see like, go do AI magic, this, this like freaks me out.

867
00:48:05.520 --> 00:48:07.400
I'm like, I don't know what's happening there.

868
00:48:07.520 --> 00:48:08.320
Do you know what I mean?

869
00:48:08.640 --> 00:48:14.680
And so like, that's, that's the smell that makes me think, okay, so now what I

870
00:48:14.680 --> 00:48:16.720
want to do is I need to unpack this.

871
00:48:17.520 --> 00:48:23.800
And if I'm unpacking, then my way of unpacking is to, is to use boxes.

872
00:48:24.280 --> 00:48:29.240
And, um, um, when it comes to the UI part, we have clear rules.

873
00:48:29.760 --> 00:48:34.280
The, the backend part of breadboarding is actually new.

874
00:48:34.600 --> 00:48:37.720
And I don't, I wouldn't say we have a clearly settled grammar yet.

875
00:48:38.520 --> 00:48:42.920
Um, uh, from like a nerdy perspective, I can tell you that everything I'm doing

876
00:48:42.920 --> 00:48:47.160
with the technical breadboarding is inspired by, um, something

877
00:48:47.160 --> 00:48:48.960
called data oriented programming.

878
00:48:49.360 --> 00:48:52.200
There's a, you, there's even a book you can find, um, from a

879
00:48:52.200 --> 00:48:53.680
friend of mine who wrote that.

880
00:48:53.880 --> 00:48:58.560
It's basically like how the functional programming world deals with software

881
00:48:58.560 --> 00:49:01.400
design, but the thing is we don't care about functional programming.

882
00:49:01.400 --> 00:49:05.360
It's just that when it comes to explaining how stuff works,

883
00:49:05.520 --> 00:49:06.880
it's the right mental model.

884
00:49:06.920 --> 00:49:08.360
So you don't have to code it that way.

885
00:49:08.640 --> 00:49:12.160
But, but from a, from a, um, understanding the system and

886
00:49:12.160 --> 00:49:16.320
reasoning about it standpoint, it's, it's, it lets you, it

887
00:49:16.320 --> 00:49:17.560
lets you clarify everything.

888
00:49:17.560 --> 00:49:22.560
So then what I'm basically doing in my head is I'm thinking about, I want

889
00:49:22.560 --> 00:49:25.080
to have some function that happens.

890
00:49:27.800 --> 00:49:32.920
So like, this is the magic AI function now.

891
00:49:32.960 --> 00:49:33.360
Right.

892
00:49:33.600 --> 00:49:34.880
But I still don't know what that means.

893
00:49:34.920 --> 00:49:36.200
So now how do I break this up?

894
00:49:36.200 --> 00:49:36.760
You know what I mean?

895
00:49:36.760 --> 00:49:43.880
So like, first there's going to be this like transcript and then there's

896
00:49:43.880 --> 00:49:49.920
going to be this like add, um, statements, you know, um, but now the

897
00:49:49.920 --> 00:49:53.320
thing is, so now there's going to be times when you feel like just

898
00:49:53.320 --> 00:49:57.600
describing the function, like you understand enough how this system

899
00:49:57.600 --> 00:49:59.880
works, that this, this is, you're like, I'm fine with that.

900
00:49:59.880 --> 00:50:00.040
I know.

901
00:50:00.000 --> 00:50:04.300
What that means, you know, um, for example, if this says like, send email,

902
00:50:04.320 --> 00:50:05.380
you're like, I know how to do that.

903
00:50:05.380 --> 00:50:06.540
We have an email service.

904
00:50:06.580 --> 00:50:07.700
I don't care how it works.

905
00:50:07.740 --> 00:50:08.140
Right.

906
00:50:08.500 --> 00:50:12.860
But if this is something which is like starting to feel weird, or we're

907
00:50:12.860 --> 00:50:15.380
going to have to work on it, we don't know how it works and that's the place

908
00:50:15.380 --> 00:50:18.860
where I'm starting to think, like, I want to describe the transformation

909
00:50:18.860 --> 00:50:24.300
that happens here and my first step of going deeper into detail is the shape

910
00:50:24.300 --> 00:50:25.860
of the input and the shape of the output.

911
00:50:25.860 --> 00:50:33.660
So I'm, I'm just using a sort of pseudo TypeScript usually for that.

912
00:50:34.100 --> 00:50:39.060
Um, and that's where I'm saying something like, if I, if I have something,

913
00:50:39.340 --> 00:50:42.660
so let's, let's say this is clean transcript.

914
00:50:42.820 --> 00:50:46.440
If I have something that we call a transcript, which is an actual object

915
00:50:46.460 --> 00:50:49.780
that means something to us as a shape, then I can say, I'm going to take in

916
00:50:49.780 --> 00:50:53.740
a transcript and then just using this sort of dirty TypeScript, like I'm going

917
00:50:53.740 --> 00:50:55.660
to say, I'm going to output a transcript as well.

918
00:50:55.740 --> 00:50:56.300
Do you know what I mean?

919
00:50:56.780 --> 00:51:00.220
So, and here's the thing where like, maybe this, we understand this, but

920
00:51:00.220 --> 00:51:01.380
maybe we don't understand this.

921
00:51:01.380 --> 00:51:03.660
So maybe we want to break this into five boxes.

922
00:51:03.700 --> 00:51:04.340
Do you know what I mean?

923
00:51:04.340 --> 00:51:05.620
That like says what happens here.

924
00:51:05.620 --> 00:51:09.340
So that's just kind of, um, where I'm at so far in terms of

925
00:51:09.340 --> 00:51:11.460
like, um, how systematized this is.

926
00:51:12.460 --> 00:51:12.740
Cool.

927
00:51:13.100 --> 00:51:16.820
Um, the one, one other thing I'll throw out is what I think, what, what, um,

928
00:51:16.860 --> 00:51:22.940
what I shared on the previous, um, check-in, um, with Lucian was there's

929
00:51:22.940 --> 00:51:26.300
times when you don't really care about the backend so much as in terms of

930
00:51:26.300 --> 00:51:32.100
processing, but you do care about the data that's coming in and out of a UI process.

931
00:51:32.420 --> 00:51:38.620
So in a case like that, we might say like, I, maybe I need to have a claim

932
00:51:38.900 --> 00:51:42.140
before I upload files because that's what the files are getting attached to.

933
00:51:42.420 --> 00:51:45.660
So then I think of this almost like a react view where like, I have a certain

934
00:51:45.660 --> 00:51:49.220
amount of like certain props that I expect to be there, you know, and then

935
00:51:49.220 --> 00:51:54.980
like at the end of this, then, um, if this upload step is happening and I have

936
00:51:54.980 --> 00:52:00.940
some kind of a save upload, then, then maybe I described the data shape of like

937
00:52:00.940 --> 00:52:03.020
what I got from that view, for example.

938
00:52:03.180 --> 00:52:03.700
Do you know what I mean?

939
00:52:03.700 --> 00:52:05.780
So this is just a little bit of like different tricks.

940
00:52:07.140 --> 00:52:07.660
Okay.

941
00:52:07.780 --> 00:52:09.100
Uh, game on.

942
00:52:10.620 --> 00:52:13.260
So we did a little bit of untangling here.

943
00:52:13.380 --> 00:52:18.220
Um, uh, where, where my head is at is like, we didn't yet talk about like what

944
00:52:18.220 --> 00:52:21.300
you guys are supposed to be changing or what is supposed to be happening.

945
00:52:21.300 --> 00:52:22.740
That's different from the current version.

946
00:52:23.180 --> 00:52:26.820
So, um, um, Bruce, do you want to get us to that?

947
00:52:28.700 --> 00:52:29.660
Yeah, this is helpful.

948
00:52:29.660 --> 00:52:34.140
I mean, I think, I mean, as far as how the current process is, it seems like

949
00:52:34.140 --> 00:52:39.700
this is a solid 30, 40% documented of what's current, um, there's still, and

950
00:52:39.700 --> 00:52:43.380
we can kind of follow unless you want to slice it to just focus on transcripts,

951
00:52:43.460 --> 00:52:46.500
like for, for what's transcripts happening currently, and then what are

952
00:52:46.780 --> 00:52:49.420
transcripts in the future, but there's this other piece of like

953
00:52:49.420 --> 00:52:51.340
the file management as well.

954
00:52:51.900 --> 00:52:52.300
Um,

955
00:52:52.900 --> 00:52:55.900
so, um, yeah, let me, so let me react to that quickly.

956
00:52:55.900 --> 00:52:59.940
And then we'll, we'll either go continue your line or we'll, or we'll drill down.

957
00:53:00.060 --> 00:53:03.900
So like in my mental map, there's this whole blob of what this thing is.

958
00:53:04.460 --> 00:53:08.700
And we have like one section of this blob is you mentioned as transcripts

959
00:53:09.100 --> 00:53:13.300
and there's some whole other, well, apparently don't even know what scale

960
00:53:13.300 --> 00:53:15.300
I'm in here.

961
00:53:16.700 --> 00:53:17.100
Okay.

962
00:53:17.420 --> 00:53:20.580
So there's, there's some part, we're enjoying watching you

963
00:53:20.580 --> 00:53:21.940
learn how to use this tool.

964
00:53:24.260 --> 00:53:28.860
And, uh, and, and what, uh, uploads was another kind of chunk here.

965
00:53:29.340 --> 00:53:29.700
Sure.

966
00:53:30.660 --> 00:53:37.180
Um, and, um, yeah, we, we actually do have, uh, maybe we can drop the monster map.

967
00:53:37.340 --> 00:53:42.100
Uh, yeah, well, this is, this is what I'm doing is this is just my, uh, uh, when I

968
00:53:42.100 --> 00:53:45.860
try to especially be, be dirty with it, like to show you how dirty it is in my

969
00:53:45.860 --> 00:53:48.420
head, like that, there's some different things here.

970
00:53:48.580 --> 00:53:52.620
So my first thought would be like, if you, if you have in your head, uh, that

971
00:53:52.620 --> 00:53:55.660
there are different things that are as part of this project, if they are

972
00:53:55.660 --> 00:54:01.540
orthogonal, we can choose the one that is the gnarliest and we can drill in.

973
00:54:02.700 --> 00:54:08.660
If they're not orthogonal and they actually have tangles between them, we

974
00:54:08.660 --> 00:54:14.020
can, we can do an untangling between the whole, the bigger areas as well.

975
00:54:14.060 --> 00:54:18.340
For example, if there's some common piece that they all depend on, or do

976
00:54:18.340 --> 00:54:18.860
you know what I mean?

977
00:54:18.860 --> 00:54:21.660
Or changing one influences how you change the other.

978
00:54:22.060 --> 00:54:29.420
Um, so, um, uh, kind of like, uh, um, what's in your head in terms of like,

979
00:54:29.940 --> 00:54:34.620
uh, is there one place for us to drill down that is clearly, um, orthogonal

980
00:54:34.620 --> 00:54:38.260
and gnarly, or is it like, there's other stuff that you don't understand it.

981
00:54:38.260 --> 00:54:39.460
And we need to pull it apart.

982
00:54:39.460 --> 00:54:41.820
Like, do you have a sense of kind of where the trouble

983
00:54:41.820 --> 00:54:43.500
is or the worst trouble is?

984
00:54:49.180 --> 00:54:56.380
Yeah, I think, um, well, I don't know necessarily that, I mean, which may speak

985
00:54:56.380 --> 00:54:59.580
to it, but I also know.

986
00:55:00.000 --> 00:55:01.580
This is all sort of, some of the stuff

987
00:55:01.580 --> 00:55:02.960
is going to sort of stay the same.

988
00:55:05.900 --> 00:55:07.340
So but we're trying to figure out

989
00:55:07.340 --> 00:55:09.660
where do the connection points from the new piece.

990
00:55:09.660 --> 00:55:11.580
So it's not like all of this is going

991
00:55:11.580 --> 00:55:14.340
to be changing in the new version.

992
00:55:14.340 --> 00:55:17.580
But yeah, what would you say is the pain, Lucian or Chris?

993
00:55:22.620 --> 00:55:24.820
LUCIANO FLORES-CHAVEZ Well, in terms of the new system,

994
00:55:24.820 --> 00:55:26.820
where we're wanting to go with this.

995
00:55:27.240 --> 00:55:28.600
Yeah, like what are the parts that

996
00:55:28.600 --> 00:55:31.120
are in your head that are like, this is something

997
00:55:31.120 --> 00:55:35.080
that we need to understand it better.

998
00:55:35.080 --> 00:55:39.800
We're wanting to not use like the summaries

999
00:55:39.800 --> 00:55:42.000
from like the database, but actually do

1000
00:55:42.000 --> 00:55:45.200
like a vector database system.

1001
00:55:45.200 --> 00:55:50.280
So it seems like for transcripts, it's like fine.

1002
00:55:50.280 --> 00:55:52.160
We'll just like grab the output and just

1003
00:55:52.160 --> 00:55:54.520
like also put it in the vector database,

1004
00:55:54.520 --> 00:55:57.300
like keep writing it where it's written now.

1005
00:55:57.300 --> 00:55:59.940
And then there's like the files.

1006
00:55:59.940 --> 00:56:03.700
So it's like a piece that we're going to put in place

1007
00:56:03.700 --> 00:56:05.500
and going to change the transcripts.

1008
00:56:05.500 --> 00:56:10.220
And then we're going to build like similar system for files.

1009
00:56:10.220 --> 00:56:13.020
But that whole thing about vector database

1010
00:56:13.020 --> 00:56:15.620
is a bit of like a unknown.

1011
00:56:15.620 --> 00:56:18.100
SIMON ULLMANN Similar system for files,

1012
00:56:18.100 --> 00:56:20.480
similar in the sense that you're going to upload the files,

1013
00:56:20.480 --> 00:56:22.100
and then you're going to chunk them out

1014
00:56:22.160 --> 00:56:24.560
and store pieces in a vector database?

1015
00:56:24.560 --> 00:56:25.520
Yeah.

1016
00:56:25.520 --> 00:56:27.640
OK, so you're basically kind of moving

1017
00:56:27.640 --> 00:56:31.440
to like a vector DB backed rag approach

1018
00:56:31.440 --> 00:56:35.200
to both the summaries and the file uploads.

1019
00:56:35.200 --> 00:56:36.280
OK.

1020
00:56:36.280 --> 00:56:38.600
It's exactly a rag agent for sure.

1021
00:56:38.600 --> 00:56:40.600
Yeah, OK.

1022
00:56:40.600 --> 00:56:42.080
OK.

1023
00:56:42.080 --> 00:56:48.860
So from an interdependency standpoint,

1024
00:56:48.860 --> 00:56:54.360
there's a do we need to?

1025
00:56:54.360 --> 00:56:57.920
OK, I'm going to we'll come back to that.

1026
00:56:57.920 --> 00:57:00.080
I don't want to go into it if we don't need it yet.

1027
00:57:00.080 --> 00:57:01.280
Let me see if we need it.

1028
00:57:01.280 --> 00:57:05.920
There's a tool for dealing with these kind of things,

1029
00:57:05.920 --> 00:57:09.880
but I don't know if we need it yet.

1030
00:57:09.880 --> 00:57:12.480
What are some other things that are in your head of like,

1031
00:57:12.480 --> 00:57:18.160
this is something else that we need to do

1032
00:57:18.160 --> 00:57:20.820
that feels like, is there anything on the level of this?

1033
00:57:20.820 --> 00:57:21.380
Do you know what I mean?

1034
00:57:21.380 --> 00:57:23.500
That's kind of like big level stuff that's in your head

1035
00:57:23.500 --> 00:57:24.500
that we're going to like other stuff

1036
00:57:24.500 --> 00:57:27.000
we're going to have to do that we're going to have to solve?

1037
00:57:30.420 --> 00:57:33.060
Yeah, we're talking about like the desired system.

1038
00:57:33.060 --> 00:57:35.860
I think the tool calling, like figuring out

1039
00:57:35.860 --> 00:57:38.300
like what tools we need for the questions

1040
00:57:38.300 --> 00:57:40.660
they're going to be typically asking

1041
00:57:40.660 --> 00:57:46.660
is like another big like, yeah, dreadful thing.

1042
00:57:49.140 --> 00:57:52.380
And where in the current system is this tool calling thing?

1043
00:57:52.380 --> 00:57:54.620
So this piece, I kind of understand where it belongs.

1044
00:57:54.620 --> 00:57:56.700
Where in the current system is this tool calling

1045
00:57:56.700 --> 00:57:59.660
thing relevant?

1046
00:57:59.660 --> 00:58:02.420
Right now, it's not because they have just

1047
00:58:02.420 --> 00:58:05.540
like preset like prompts for like background processing.

1048
00:58:05.540 --> 00:58:09.140
And this tool calling is like a new piece for like the chat.

1049
00:58:09.140 --> 00:58:12.940
Like they're going to be just chatting with in the app.

1050
00:58:12.940 --> 00:58:17.340
And so there's a so there if I'm coming back

1051
00:58:17.360 --> 00:58:24.280
to my mental map here, like there's this chat.

1052
00:58:24.280 --> 00:58:25.960
There's the chat is a different piece

1053
00:58:25.960 --> 00:58:27.480
of this that's different from the file upload

1054
00:58:27.480 --> 00:58:28.360
and the transcripts.

1055
00:58:28.360 --> 00:58:30.320
OK.

1056
00:58:30.320 --> 00:58:35.960
Is it safe to say like it sounds like the chat

1057
00:58:35.960 --> 00:58:39.680
is kind of orthogonal to the transcripts and the files,

1058
00:58:39.680 --> 00:58:41.240
but the transcripts and the files

1059
00:58:41.240 --> 00:58:45.880
feel less orthogonal because they both

1060
00:58:45.880 --> 00:58:49.580
have to do with kind of moving to this vector DB

1061
00:58:49.580 --> 00:58:51.860
approach to how you're storing stuff so that you can look it

1062
00:58:51.860 --> 00:58:52.360
up.

1063
00:58:52.360 --> 00:58:53.740
Is that fair to say?

1064
00:58:53.740 --> 00:58:55.140
Sounds fair, yeah.

1065
00:58:55.140 --> 00:58:56.540
OK.

1066
00:58:56.540 --> 00:59:00.460
And if transcripts and files don't happen,

1067
00:59:00.460 --> 00:59:04.060
then there's nothing interesting for your tools to do in chat.

1068
00:59:04.060 --> 00:59:05.620
Is that right?

1069
00:59:05.620 --> 00:59:06.460
Yeah.

1070
00:59:06.460 --> 00:59:07.100
OK.

1071
00:59:07.100 --> 00:59:11.260
So the mental picture that I just kind of went through,

1072
00:59:11.280 --> 00:59:18.440
this is we don't need to go far into this,

1073
00:59:18.440 --> 00:59:28.240
but there's a very, very short version is there's the if we

1074
00:59:28.240 --> 00:59:30.080
factor out vectors from both of these,

1075
00:59:30.080 --> 00:59:32.800
we have like how the hell do we implement a vector

1076
00:59:32.800 --> 00:59:35.360
DB if we haven't done it before.

1077
00:59:35.360 --> 00:59:39.240
So I'm going to write like vector DB as a thing.

1078
00:59:39.240 --> 00:59:39.920
OK.

1079
00:59:39.940 --> 00:59:51.540
Then there's process files with vectors.

1080
00:59:51.540 --> 00:59:57.700
There's move summary storage to vector.

1081
01:00:00.000 --> 01:00:09.800
The interdependence here in terms of the question that I'm asking is, I'm trying to identify separate pieces and then I'm asking, what helps me do what?

1082
01:00:11.200 --> 01:00:13.800
What's the thing that I do that gives me the other thing?

1083
01:00:14.600 --> 01:00:19.600
You know, so like VectorDB helps me to do...

1084
01:00:20.800 --> 01:00:21.600
My goodness.

1085
01:00:23.800 --> 01:00:27.200
We should send this to the Figma people, I think, and then they can...

1086
01:00:30.400 --> 01:00:31.200
They can...

1087
01:00:34.400 --> 01:00:39.600
Because when you try to move fast, man, everything gets harder than when you're doing a beautiful demo for your team.

1088
01:00:39.600 --> 01:00:40.000
Okay.

1089
01:00:41.000 --> 01:00:42.400
This helps me...

1090
01:00:44.800 --> 01:00:46.200
My goodness, please just...

1091
01:00:49.600 --> 01:00:50.200
Let's just...

1092
01:00:50.800 --> 01:00:54.400
Sorry, this is going to make more sense if we do like this.

1093
01:00:56.200 --> 01:00:58.800
These help me do this, this helps me do this, and then...

1094
01:01:00.600 --> 01:01:04.800
Like, I'm going to...

1095
01:01:05.000 --> 01:01:05.800
This is going to...

1096
01:01:06.800 --> 01:01:07.400
I'm going to feel...

1097
01:01:07.400 --> 01:01:13.600
I'm probably going to feel better doing these things if these things are already there for me to call upon.

1098
01:01:15.900 --> 01:01:17.900
Does this resonate with how you're thinking?

1099
01:01:19.400 --> 01:01:20.100
100%.

1100
01:01:20.400 --> 01:01:20.900
100%?

1101
01:01:20.900 --> 01:01:21.500
Okay, good.

1102
01:01:22.500 --> 01:01:25.600
So, this is our...

1103
01:01:26.200 --> 01:01:28.200
This is where the true...

1104
01:01:29.800 --> 01:01:31.700
This is the bad place here.

1105
01:01:31.700 --> 01:01:32.400
This is the...

1106
01:01:35.200 --> 01:01:36.100
This is like...

1107
01:01:37.000 --> 01:01:38.900
I mean, because if we don't do this, then...

1108
01:01:39.900 --> 01:01:40.500
You know what I mean?

1109
01:01:40.500 --> 01:01:43.000
Everything is downstream from this, basically, right?

1110
01:01:43.400 --> 01:01:51.500
Do you have a sense of whether one of these feels like a bigger win?

1111
01:01:52.000 --> 01:02:04.000
If you were to go to the client or just for yourselves to say like, oh, summaries are now in the VectorDB versus processing of files, processing of uploaded files or Vector, does it make any difference to you?

1112
01:02:04.500 --> 01:02:08.600
Does one of them feel like a bigger win?

1113
01:02:09.500 --> 01:02:16.000
Probably files, because like the project name is like case GPT chat with files.

1114
01:02:17.200 --> 01:02:18.000
Uh-huh.

1115
01:02:18.000 --> 01:02:18.600
Okay.

1116
01:02:18.700 --> 01:02:19.200
Uh-huh.

1117
01:02:19.700 --> 01:02:20.200
Well, yeah.

1118
01:02:20.200 --> 01:02:23.900
So, we're currently not being summarized, like transcripts are.

1119
01:02:25.100 --> 01:02:27.000
So, it feels like a bigger piece, like we'll...

1120
01:02:27.100 --> 01:02:27.800
Right.

1121
01:02:28.100 --> 01:02:37.000
So, this is a little bit more like a refactoring, like you're changing the implementation versus here you're doing something new that's not possible before, right?

1122
01:02:37.500 --> 01:02:40.200
So, I'm just chunking this out into...

1123
01:02:40.300 --> 01:02:42.100
This is kind of how I'm getting to the...

1124
01:02:42.100 --> 01:02:46.000
This is kind of the structure of what's really like in the sort of nine boxes.

1125
01:02:46.000 --> 01:02:46.700
You know what I mean?

1126
01:02:47.200 --> 01:02:48.200
And it's like, okay.

1127
01:02:48.200 --> 01:02:56.300
So, this is first order of business, but we can't do this in isolation because we have to have something we're trying to actually work on.

1128
01:02:56.300 --> 01:02:57.000
Do you know what I mean?

1129
01:02:57.200 --> 01:02:58.600
So, we kind of take this together.

1130
01:02:58.800 --> 01:02:59.600
This is funny.

1131
01:02:59.600 --> 01:03:00.300
What is this?

1132
01:03:00.900 --> 01:03:02.300
If you wave your mouse, I think...

1133
01:03:05.100 --> 01:03:06.700
I never noticed that right now.

1134
01:03:07.400 --> 01:03:08.600
I don't know how that's useful.

1135
01:03:09.100 --> 01:03:09.700
It's funny.

1136
01:03:14.400 --> 01:03:15.200
So, okay.

1137
01:03:15.200 --> 01:03:21.300
So, my instinct is that like our first order of business now is to dive into this question.

1138
01:03:22.400 --> 01:03:22.900
Yeah.

1139
01:03:29.300 --> 01:03:30.600
I can pause here.

1140
01:03:33.700 --> 01:03:35.500
Am I just telling you what you guys already knew?

1141
01:03:36.100 --> 01:03:37.100
No, this is great.

1142
01:03:40.000 --> 01:03:44.200
So, this part is super useful.

1143
01:03:44.200 --> 01:03:51.100
Actually, it's like clarifies a lot because I had like kind of like similar order of business.

1144
01:03:51.100 --> 01:03:53.600
They're like on the same like scale, like tool calling.

1145
01:03:53.600 --> 01:03:57.900
Like, I thought like, okay, like figuring out what tools to call.

1146
01:03:58.200 --> 01:04:01.400
It's going to be like, but like the vector that actually clarifies a lot.

1147
01:04:01.500 --> 01:04:08.300
If I just focus on vectors and like putting the files in there, like calling the database,

1148
01:04:10.100 --> 01:04:15.200
they'll like solve a lot of like the angst, the unknowns.

1149
01:04:15.700 --> 01:04:16.100
Got it.

1150
01:04:16.100 --> 01:04:16.600
Okay.

1151
01:04:17.000 --> 01:04:17.400
Yeah.

1152
01:04:17.400 --> 01:04:18.000
Sorry, Chris.

1153
01:04:19.000 --> 01:04:23.200
I was just saying, you know, I was wondering how sometimes you get to that chunk, right?

1154
01:04:23.200 --> 01:04:23.700
And...

1155
01:04:23.700 --> 01:04:24.200
Here?

1156
01:04:24.800 --> 01:04:29.000
Yeah, the skeleton kind of what you build on for the chunk.

1157
01:04:29.700 --> 01:04:31.600
That was sort of a missing piece in my brain too.

1158
01:04:31.600 --> 01:04:32.800
So, this is super helpful.

1159
01:04:32.900 --> 01:04:33.400
Yeah.

1160
01:04:33.400 --> 01:04:33.900
Cool.

1161
01:04:33.900 --> 01:04:34.400
Okay.

1162
01:04:34.800 --> 01:04:40.300
I'm going to give you guys the tool of how I learned to do this outside of my head.

1163
01:04:42.000 --> 01:04:47.300
I mean, it's also in my head, but I feel like the way that we could like explain it.

1164
01:04:47.700 --> 01:04:54.300
This is something that I noticed that like really badass developers were doing in their minds.

1165
01:04:55.200 --> 01:04:56.700
And it's kind of like, how did you do that?

1166
01:04:56.700 --> 01:04:59.400
And I could never explain it until I learned this.

1167
01:05:00.000 --> 01:05:07.440
like a little tool. So I learned this tool from Bob. He calls it the interdependency diagram

1168
01:05:08.560 --> 01:05:12.640
and it's... is there such a thing as transparency? I mean, how do I...

1169
01:05:12.640 --> 01:05:16.240
There is. If you click on the... I think if you click up

1170
01:05:18.720 --> 01:05:20.640
here, like on this guy, then it...

1171
01:05:20.640 --> 01:05:22.480
Ah, this guy. Yeah, okay.

1172
01:05:23.280 --> 01:05:23.780
Transparent.

1173
01:05:25.600 --> 01:05:33.040
No, Phil. Okay, good. Here's the principle. I have a whole bunch of... so what I do is I have

1174
01:05:33.040 --> 01:05:38.240
a whole bunch of things that I know I need to do, okay? And I'm just gonna... and we do it

1175
01:05:38.240 --> 01:05:46.800
first as a list. So I need to process files with vectors, you know? I need to do tool calling

1176
01:05:46.800 --> 01:05:53.680
in chat. I need to... whoops. I need to move summary storage to vector.

1177
01:05:56.240 --> 01:06:04.960
And then the thing that I did in my head here was I factored out the common... it's just like math,

1178
01:06:04.960 --> 01:06:10.320
you know what I mean? You see... wait a minute. You know what I mean? There's a piece that both

1179
01:06:10.320 --> 01:06:17.440
of these things use. Therefore, like, if I know how to do that, if this was like save information

1180
01:06:17.440 --> 01:06:21.840
to database, I'm not going to pull database out as an unknown because I know how to do that, right?

1181
01:06:22.800 --> 01:06:32.240
But here, figure out vector db, you know, mechanics is a separate thing. So that's...

1182
01:06:32.240 --> 01:06:38.960
this was the first step that I did was I just got to... here seems to be the sort of list of things

1183
01:06:38.960 --> 01:06:45.920
that we kind of have to solve. And then the interdependency diagram basically works like this.

1184
01:06:45.920 --> 01:06:51.200
The idea is you take... and you can do this on anything. You put the things around a circle

1185
01:06:52.160 --> 01:06:58.720
in arbitrary positions. And here we just have four. And then you simply ask for each one.

1186
01:06:59.440 --> 01:07:06.400
There's the algorithm is like, does this help me do any of the rest? And you do it for each one.

1187
01:07:06.400 --> 01:07:13.200
So does summary storage in db help me with tool calling?

1188
01:07:16.480 --> 01:07:22.240
And in this case, it kind of doesn't because we already have summaries, right?

1189
01:07:25.120 --> 01:07:31.440
So, okay, not relevant. Does it help me with figuring out the vector db? Well, actually, I need to

1190
01:07:31.440 --> 01:07:38.000
figure out the vector db in order to do this. So, no. If I do this, does it help me to process files?

1191
01:07:38.000 --> 01:07:42.560
Well, like, only if I learn about the vector db. But then I come here to vector db and it's like, yes,

1192
01:07:42.560 --> 01:07:48.480
this helps me do this. Yes, this helps me do this. Does this directly help me do tool calling?

1193
01:07:49.280 --> 01:07:53.760
Well, not directly. I wouldn't say that I'm going to be able to do tool calling when I finish that.

1194
01:07:53.760 --> 01:07:59.200
But we could even say that if we wanted to, yes, we could. If we say that this is like...

1195
01:07:59.200 --> 01:08:03.920
There's always a question of your interpretation. But then you're going to... You could just go

1196
01:08:03.920 --> 01:08:07.200
around the circle and say for each one of these things, does it help me or does it not?

1197
01:08:07.200 --> 01:08:12.320
And what you end up with is the equivalent of this graph. You know what I mean? It's how you

1198
01:08:12.320 --> 01:08:18.160
end up with the fact that, like, a vector is on top. It's like it's got the most connection points.

1199
01:08:18.160 --> 01:08:24.560
Like, exactly. It's a bottleneck, yeah. What you end up doing algorithmically and is you... So, like, this

1200
01:08:24.560 --> 01:08:29.359
is going to help me with tool calling. It's actually funny because now that we talked about it,

1201
01:08:29.359 --> 01:08:35.279
it sounds like this might not really help me with tool calling because I already have it.

1202
01:08:35.279 --> 01:08:39.600
So, that's an interesting... It's kind of more and more argument that this is less... This is kind of

1203
01:08:39.600 --> 01:08:45.760
falling off a little bit. You know what I mean? And then when you do this on a complex problem,

1204
01:08:45.760 --> 01:08:54.640
you will have, like, a whole web of things. And then what you can do is you can literally count

1205
01:08:54.640 --> 01:08:59.120
the inputs and outputs of each thing. And then you sequence by that. And the things that have

1206
01:08:59.120 --> 01:09:03.359
more outputs than inputs float to the top of your sequence. And it tells you what order to solve

1207
01:09:03.359 --> 01:09:09.760
problems in, basically. So, that's the interdependency diagram tool.

1208
01:09:09.760 --> 01:09:13.520
It's a little hard to see it because with these four, it's so straightforward. But when you have a

1209
01:09:13.520 --> 01:09:25.680
harder problem, you can do this and you get... I have one post that I could...

1210
01:09:25.680 --> 01:09:37.680
Let me just see if I have it here.

1211
01:09:41.040 --> 01:09:45.120
No. Okay. If I find it, I'll send it to you guys. I have a post that kind of walks through.

1212
01:09:45.680 --> 01:09:50.640
Might be...

1213
01:09:57.040 --> 01:10:04.640
Yeah, here we go. Interrelate.

1214
01:10:00.000 --> 01:10:08.160
Oh that's funny. Okay. I have a post for it somewhere. We'll find it. I'll send it to you.

1215
01:10:08.160 --> 01:10:11.120
Then you'll have just a better reference in case you want to be able to do that. But okay, I don't

1216
01:10:11.120 --> 01:10:18.640
want to take up too much time. So we arrived at this is the hard thing. So then

1217
01:10:19.600 --> 01:10:25.120
what I want to be able to do is find the boundaries in the current system of where

1218
01:10:25.120 --> 01:10:28.240
this thing needs to happen and that's going to help us to create test boundaries.

1219
01:10:29.600 --> 01:10:37.760
So if I need to go research VectorDB in a pure science project,

1220
01:10:37.760 --> 01:10:42.240
that's different than if I need to research how to do a VectorDB for this specific problem.

1221
01:10:42.800 --> 01:10:51.840
So I'm just kind of wondering, can we isolate where this step would happen? So where my

1222
01:10:51.840 --> 01:11:00.080
head goes is I'm going to go to process files with VectorDB. This is here.

1223
01:11:04.560 --> 01:11:13.120
And so now the question is, all I want to do is I'm comfortable with a black box of vector magic.

1224
01:11:14.080 --> 01:11:19.200
You know, and like, we don't know what this is, but if I can have inputs to this and outputs to

1225
01:11:19.200 --> 01:11:26.480
that, that's going to help us, right? Like that's going to bound this, that, this, this, this, the,

1226
01:11:26.480 --> 01:11:33.120
this is the, we're, we're, we're very clearly bounding the hole where the research needs to

1227
01:11:33.120 --> 01:11:37.120
happen. And that's going to make the research feel much more focused and small.

1228
01:11:37.120 --> 01:11:45.760
Right. That makes sense. And if I can, just one context, Luchin, like how much of the VectorDB

1229
01:11:45.760 --> 01:11:51.120
stuff, the one missing piece of this is there's Chris Luchin and there's MJ who's, who has built

1230
01:11:51.120 --> 01:11:55.840
these rag agents before. And so he came on specifically to help us with this. I was just

1231
01:11:55.840 --> 01:12:00.400
curious, like how much of this is he solving versus you? Has he solved it? I know he's been

1232
01:12:00.400 --> 01:12:05.760
working on a bunch of stuff, but how much of this is he solving versus you? Has he solved it? I know

1233
01:12:05.840 --> 01:12:09.280
he's been working on a bunch of stuff. Yeah. He gave me a template. Like he wrote

1234
01:12:11.120 --> 01:12:19.200
basically a bunch of scripts that take like files and like put it in the, in the VectorDB and like,

1235
01:12:21.600 --> 01:12:25.840
like templates for like the tool calling to actually get the data from the VectorDB.

1236
01:12:28.640 --> 01:12:32.960
I started translating that like from Python to JavaScript and like just kind of putting it in

1237
01:12:32.960 --> 01:12:38.160
place. But just with like all the different things, I just get like way more clarity now

1238
01:12:38.160 --> 01:12:42.960
that this is like the central piece to focus on. So I just kind of put it to the side for a bit.

1239
01:12:43.760 --> 01:12:47.440
Yeah. I just, I was, the red flag for me was that like, you're trying to solve a problem that we

1240
01:12:47.440 --> 01:12:52.880
specifically got him to help come in and not just give you a template. He was supposed to come in

1241
01:12:52.880 --> 01:12:59.680
and sort of help you build it. And then, yeah, this seems like this. I figured like he gave me

1242
01:12:59.680 --> 01:13:03.280
everything that I needed so I can just like kind of wrap up the rest. Yeah.

1243
01:13:04.720 --> 01:13:14.000
So here's a place where if, if we take this, this this, there's some VectorDB work.

1244
01:13:15.120 --> 01:13:21.840
If we can just understand where this lives and, and, and what goes in and what comes out of it,

1245
01:13:22.800 --> 01:13:26.800
then this will, should feel almost like a mini project.

1246
01:13:28.880 --> 01:13:33.920
And when it's a mini project, Bruce, you guys can have the conversation of like, almost like,

1247
01:13:33.920 --> 01:13:38.720
you can think of it almost like who's on that project. You know what I mean? And like you can

1248
01:13:38.720 --> 01:13:43.760
have a very, very clear conversation around like sort of who's responsible versus who's advising

1249
01:13:43.760 --> 01:13:50.480
or, you know what I mean? Like that kind of a thing. So is this, is this a simple answer

1250
01:13:51.120 --> 01:13:56.000
of like where this fits into the current system and where it happens and what,

1251
01:13:56.000 --> 01:14:01.680
and what the outputs are and, and like kind of where the seams are in the system?

1252
01:14:04.240 --> 01:14:12.720
I think so. Yeah. Just any file that's like uploaded, just get summary. We can like tweak

1253
01:14:12.720 --> 01:14:18.960
the prompt for getting something and we just save that. And then just like tool calling of like,

1254
01:14:18.960 --> 01:14:24.720
Hey, I need like this question answered about like using files.

1255
01:14:27.920 --> 01:14:35.440
So like high level, but to me, I don't feel like too concerned about that. It feels like it doesn't

1256
01:14:35.440 --> 01:14:45.280
need too much more resolution. Okay. So I'm, I'm I'm inclined to delete this black box. I don't

1257
01:14:45.280 --> 01:14:53.280
feel like we need to have this like intentionally scary unknown marked out. What I'm hearing is

1258
01:14:54.080 --> 01:14:59.760
there is a when does that happen?

1259
01:15:00.000 --> 01:15:02.920
So the S3 upload happens, right?

1260
01:15:02.920 --> 01:15:12.040
And where does this processing begin of this new step

1261
01:15:12.040 --> 01:15:14.880
where we're going to do the prompt and all of this?

1262
01:15:18.560 --> 01:15:20.240
I'm not sure I understand the question.

1263
01:15:20.240 --> 01:15:20.880
OK.

1264
01:15:20.880 --> 01:15:25.080
So there's a new step that needs to happen

1265
01:15:25.080 --> 01:15:32.160
where the uploaded file, something

1266
01:15:32.160 --> 01:15:35.040
needs to happen to it so it has a representation in the vector

1267
01:15:35.040 --> 01:15:36.480
DB.

1268
01:15:36.480 --> 01:15:37.720
Is that right?

1269
01:15:37.720 --> 01:15:38.400
Yeah.

1270
01:15:38.400 --> 01:15:40.440
Yeah.

1271
01:15:40.440 --> 01:15:43.200
So maybe it sounds like a stupid question,

1272
01:15:43.200 --> 01:15:47.680
but when does that process start?

1273
01:15:47.680 --> 01:15:50.720
So immediately after the file is uploaded,

1274
01:15:50.720 --> 01:15:54.600
is this when a job is enqueued or it just

1275
01:15:54.600 --> 01:15:57.160
directly something gets fired?

1276
01:15:57.160 --> 01:16:00.120
Yeah, I would do it that way.

1277
01:16:00.120 --> 01:16:00.760
OK.

1278
01:16:00.760 --> 01:16:06.680
So when, so here's one other thing we can do.

1279
01:16:06.680 --> 01:16:14.520
I'm just going to say we are going to add job to queue

1280
01:16:14.520 --> 01:16:19.920
to process or to even to vectorize.

1281
01:16:19.920 --> 01:16:33.560
And then when the vectorized job gets triggered,

1282
01:16:33.560 --> 01:16:36.040
so at some time this is going to get triggered,

1283
01:16:36.040 --> 01:16:39.280
then you just gave a high level.

1284
01:16:39.280 --> 01:16:40.160
What needs to happen?

1285
01:16:45.280 --> 01:16:48.880
So I've got an uploaded file in S3,

1286
01:16:48.880 --> 01:16:50.920
and I'm supposed to be vectorizing it.

1287
01:16:50.920 --> 01:16:52.760
What's the first thing that needs to happen?

1288
01:16:56.320 --> 01:17:01.400
Extract the contents of the file itself.

1289
01:17:01.400 --> 01:17:02.560
This could be a rabbit hole.

1290
01:17:02.560 --> 01:17:05.240
It's unknown how much because some of them

1291
01:17:05.240 --> 01:17:09.280
might be just require OCR.

1292
01:17:09.280 --> 01:17:10.280
Some of them are fine.

1293
01:17:10.280 --> 01:17:13.480
I can just use the existing libraries.

1294
01:17:13.480 --> 01:17:14.120
Uh-huh.

1295
01:17:14.120 --> 01:17:14.620
OK.

1296
01:17:18.880 --> 01:17:20.440
This other one, I'm not sure.

1297
01:17:20.440 --> 01:17:26.160
It's like seems optional, but like a clean up?

1298
01:17:26.160 --> 01:17:29.440
Yeah, probably not optional.

1299
01:17:29.440 --> 01:17:30.440
What does clean up mean?

1300
01:17:33.440 --> 01:17:37.480
A bunch of lines and like dash, dash, dash, dash, dash

1301
01:17:37.480 --> 01:17:40.120
to like signify like a just divider

1302
01:17:40.120 --> 01:17:42.480
or a bunch of like useless stuff.

1303
01:17:42.480 --> 01:17:43.280
Formatting maybe?

1304
01:17:43.280 --> 01:17:43.780
Uh-huh.

1305
01:17:44.780 --> 01:17:50.740
Um, I think one thing we said as well was,

1306
01:17:50.740 --> 01:17:52.940
I can't remember if it was this project or another one,

1307
01:17:52.940 --> 01:17:56.060
where we did say we would limit it to,

1308
01:17:56.060 --> 01:17:57.580
because you said like the OCR stuff.

1309
01:17:57.580 --> 01:18:00.460
Like I think we said we would just basically keep it to,

1310
01:18:00.460 --> 01:18:01.260
is it this project?

1311
01:18:01.260 --> 01:18:03.020
I may be getting them confused.

1312
01:18:03.020 --> 01:18:05.300
But there was a different formatting

1313
01:18:05.300 --> 01:18:08.100
where like it's basically a readable PDF.

1314
01:18:08.100 --> 01:18:10.380
Or is that for the, I think it's the file filling out

1315
01:18:10.380 --> 01:18:11.620
thing, the next project.

1316
01:18:11.620 --> 01:18:12.220
Yeah.

1317
01:18:12.220 --> 01:18:15.260
Yeah, that's like a huge rabbit hole for another project.

1318
01:18:15.260 --> 01:18:18.740
So this is still a valid question for this project.

1319
01:18:18.740 --> 01:18:22.380
Like how many of these files are like difficult to read?

1320
01:18:22.380 --> 01:18:24.220
Yeah, I think that's a good boundary

1321
01:18:24.220 --> 01:18:26.260
to draw somewhere there, because that probably

1322
01:18:26.260 --> 01:18:27.420
is an infinite rabbit hole.

1323
01:18:27.420 --> 01:18:29.000
Another thing that we're going to hear

1324
01:18:29.000 --> 01:18:31.540
is like putting in like a markdown format

1325
01:18:31.540 --> 01:18:33.580
and having like proper headings.

1326
01:18:33.580 --> 01:18:36.660
So knowing like which questions they'll ask

1327
01:18:36.660 --> 01:18:43.060
will be like I think relevant to like summarizing

1328
01:18:43.060 --> 01:18:44.220
or like structuring.

1329
01:18:44.220 --> 01:18:48.740
I think it's a better word, like structuring the contents.

1330
01:18:48.740 --> 01:18:50.220
So did I hear that right?

1331
01:18:50.220 --> 01:18:55.100
That the questions they want to ask from the chat tooling

1332
01:18:55.100 --> 01:18:58.140
are influencing how you're going to extract the contents?

1333
01:19:00.860 --> 01:19:04.820
Yeah, and we have a list of those desired prompts

1334
01:19:04.820 --> 01:19:06.860
that they would love to eventually ask.

1335
01:19:06.860 --> 01:19:10.020
Like, what's the next action to move this claim forward?

1336
01:19:10.020 --> 01:19:11.700
Like it would need to know a lot more.

1337
01:19:11.700 --> 01:19:15.460
Like what contents are missing on this claim,

1338
01:19:15.460 --> 01:19:17.900
or stuff like that.

1339
01:19:17.900 --> 01:19:21.420
OK, I'm going to just do something here real quick.

1340
01:19:21.420 --> 01:19:32.900
I'm going to, so I'm just making a top level thing here

1341
01:19:32.900 --> 01:19:34.420
for all of this.

1342
01:19:35.300 --> 01:19:37.900
I just pulled this out into saying this gets triggered.

1343
01:19:37.900 --> 01:19:47.620
And I think something that's interesting here is,

1344
01:19:47.620 --> 01:19:49.740
I don't know how to call this out exactly.

1345
01:19:49.740 --> 01:19:57.300
But what I'm just noting is the expected prompts

1346
01:19:57.300 --> 01:19:59.860
are an input to you when you're designing this.

1347
01:20:00.000 --> 01:20:09.120
Like, when you are trying to solve how to extract contents so that it works well, you're

1348
01:20:09.120 --> 01:20:13.560
referring to the expected prompts, okay.

1349
01:20:13.560 --> 01:20:18.880
Yeah, but it's not actually an input here, so okay, that's fine, that's just something

1350
01:20:18.880 --> 01:20:22.200
that we can keep in mind.

1351
01:20:22.200 --> 01:20:25.200
It's almost like I kind of have a comment in my head here, but I don't know how to put

1352
01:20:25.200 --> 01:20:26.280
it here, you know what I mean?

1353
01:20:26.280 --> 01:20:38.200
It's kind of like, yeah, that would be more for like formatting, for like extracting content

1354
01:20:38.200 --> 01:20:41.280
that's pretty standard thing, like we.

1355
01:20:41.280 --> 01:20:42.280
That's more for formatting, right?

1356
01:20:42.280 --> 01:20:46.960
It's the file as the input, and just the super launch, and that's the output.

1357
01:20:46.960 --> 01:20:51.280
Is cleanup different than formatting?

1358
01:20:51.280 --> 01:20:55.040
I think so, yeah, formatting you can think of like a markdown file with like different

1359
01:20:55.040 --> 01:21:00.400
headings and sections.

1360
01:21:00.400 --> 01:21:07.840
And I would add another, before it like slips my mind, but like determining like file type,

1361
01:21:07.840 --> 01:21:08.840
right?

1362
01:21:08.840 --> 01:21:13.760
Because some files are like medical evidence, some files are like.

1363
01:21:13.760 --> 01:21:16.760
That's this here, yeah?

1364
01:21:16.760 --> 01:21:18.880
Yeah, yeah.

1365
01:21:18.880 --> 01:21:25.560
I think they'll be at the tool calling level, like to know like which type of files to query.

1366
01:21:25.560 --> 01:21:30.580
This will be at the tool calling level to know which type of files to query.

1367
01:21:30.580 --> 01:21:34.600
When you say that, are you referring to the tool calling from the chat after this has

1368
01:21:34.600 --> 01:21:35.600
all happened?

1369
01:21:35.600 --> 01:21:36.600
Yeah.

1370
01:21:36.600 --> 01:21:38.360
Yeah, I'm kind of going backwards.

1371
01:21:38.360 --> 01:21:41.160
What are the like expected inputs to the chat?

1372
01:21:41.160 --> 01:21:42.160
Cool.

1373
01:21:42.160 --> 01:21:43.160
Yeah, yeah.

1374
01:21:43.160 --> 01:21:45.720
That's what I, the output from here.

1375
01:21:45.720 --> 01:21:46.720
Yeah.

1376
01:21:46.720 --> 01:21:53.160
That's interesting how thinking about chat is actually helping us to do this.

1377
01:21:53.160 --> 01:21:55.320
Chat is like centerpiece.

1378
01:21:55.320 --> 01:21:57.400
That's the whole point of doing this.

1379
01:21:57.400 --> 01:21:59.160
Yeah, that's interesting.

1380
01:21:59.160 --> 01:22:05.800
I'm kind of updating my mental model here where it's almost like I want to have a, I

1381
01:22:05.800 --> 01:22:13.040
want to have like a, so like in my mental model now, it's kind of like there's a, I

1382
01:22:13.040 --> 01:22:19.840
want to at least have an, I want to have like a, I want to at least have like an understanding

1383
01:22:19.840 --> 01:22:32.080
of tool calling in chat as an input to how we do this X, to how we do this processing

1384
01:22:32.080 --> 01:22:33.080
here.

1385
01:22:33.080 --> 01:22:34.800
You know what I mean?

1386
01:22:34.800 --> 01:22:39.240
It's like, that's another thing, like we're not doing like, we're not doing this completely

1387
01:22:39.240 --> 01:22:40.240
without any understanding.

1388
01:22:40.440 --> 01:22:43.160
Like you're actually trying to think about like what needs to be happening in the tool

1389
01:22:43.160 --> 01:22:44.560
calling as an input to this.

1390
01:22:44.560 --> 01:22:48.160
And then this is like tool calling actually works.

1391
01:22:48.160 --> 01:22:49.840
You know what I mean?

1392
01:22:49.840 --> 01:23:00.320
There's a difference between kind of, well, yeah, we need to have, we do need to have

1393
01:23:00.320 --> 01:23:02.200
like some sense of what tool calling means.

1394
01:23:02.200 --> 01:23:10.000
Sorry, this is just like, this is killing me, these, these, these arrows here.

1395
01:23:10.640 --> 01:23:16.760
What tool calling itself is that I feel is pretty clear, but what is not clear is like,

1396
01:23:16.760 --> 01:23:21.120
which tools do we need for the chat to be able to call?

1397
01:23:21.120 --> 01:23:22.120
Uh-huh.

1398
01:23:22.120 --> 01:23:23.120
Same one.

1399
01:23:23.120 --> 01:23:26.040
I think that it's still a valid point.

1400
01:23:26.040 --> 01:23:29.360
Like tool calling, I don't know how familiar you are with it.

1401
01:23:29.360 --> 01:23:31.760
So I'll just give you like the footnotes.

1402
01:23:31.760 --> 01:23:35.480
When I think of tool calling, I think of like cloud code deciding that it's going to use

1403
01:23:35.480 --> 01:23:40.640
grep versus it's going to run some tests or versus something like that.

1404
01:23:40.640 --> 01:23:41.640
No.

1405
01:23:41.640 --> 01:23:50.720
I think Lucian just lost connectivity for a sec.

1406
01:23:50.720 --> 01:23:51.720
It may.

1407
01:23:51.720 --> 01:23:52.720
Yeah.

1408
01:23:52.720 --> 01:23:53.720
I'll be back.

1409
01:23:53.720 --> 01:24:01.480
Um, I did grab, um, the, uh, the, just the chat interface of where that sits is sitting

1410
01:24:01.480 --> 01:24:08.040
in the kind of often over to the, I can bring it over to you if you want, because there's

1411
01:24:08.040 --> 01:24:14.040
this interface is sort of what questions you can ask in this option B, I think, uh, towards

1412
01:24:14.040 --> 01:24:16.040
the bottom left.

1413
01:24:16.040 --> 01:24:17.040
Yeah.

1414
01:24:17.040 --> 01:24:23.200
Um, this, this does kind of, um, this piece right here is kind of where the, this is sort

1415
01:24:23.200 --> 01:24:25.260
of the interface piece of what we're building.

1416
01:24:25.260 --> 01:24:30.960
So there's going to be a chat, um, detail page, uh, where you click on a chat, you know,

1417
01:24:30.960 --> 01:24:36.520
like chat GPT sort of style, um, and that, uh, you would see your past conversations.

1418
01:24:36.520 --> 01:24:39.680
You would see, you could start a new chat, the new chat would, you know, have these pieces

1419
01:24:39.680 --> 01:24:43.880
and that's where it would then, when you send the message, then it's start kicks off this

1420
01:24:43.880 --> 01:24:44.880
whole process.

1421
01:24:44.880 --> 01:24:48.920
So we started like building out, you know, what happens with this, there's sort of a

1422
01:24:48.920 --> 01:24:50.960
default prompt that gets included.

1423
01:24:50.960 --> 01:24:55.400
There's like, um, you know, there's going to be other contexts that's added.

1424
01:24:55.400 --> 01:25:00.000
Then we take all of that and we kick it over to, to the open AI.

1425
01:25:00.000 --> 01:25:01.020
Which is going to then tool it.

1426
01:25:01.080 --> 01:25:04.620
And then that's where you started building out these tools here, like

1427
01:25:04.620 --> 01:25:08.220
list files in claim, like, Oh, we have these 10 files, but we need these three

1428
01:25:08.220 --> 01:25:11.220
other three, so that's like one tool potentially that could be used.

1429
01:25:11.700 --> 01:25:12.300
I see.

1430
01:25:13.660 --> 01:25:19.640
Um, and this is a, um, this is like a tool, um, uh, in the sense of like, is

1431
01:25:19.640 --> 01:25:23.660
the tool actually just another prompt, um, that's kind of named that, that, that,

1432
01:25:24.100 --> 01:25:25.580
uh, okay, got it.

1433
01:25:27.440 --> 01:25:35.920
Um, um, so it's like, do you, do you, I'm sorry.

1434
01:25:35.920 --> 01:25:36.560
What was that?

1435
01:25:37.820 --> 01:25:41.120
Uh, I don't know if this is nitpicking or it's actually useful, but these are

1436
01:25:41.120 --> 01:25:45.540
like functions, like for example, if you ask Chad GPT, like, Hey, what's the

1437
01:25:45.540 --> 01:25:50.320
weather today, there's like a list of like, you can fetch this API to answer

1438
01:25:50.320 --> 01:25:53.020
this and then put it back in like human readable.

1439
01:25:53.080 --> 01:25:56.680
Uh, because this is actually something you're going to go via, you're going to

1440
01:25:56.680 --> 01:26:00.560
go back to the actual database and like fetch stuff that treachery needs to know.

1441
01:26:00.580 --> 01:26:01.060
I got it.

1442
01:26:01.080 --> 01:26:01.480
Okay.

1443
01:26:01.540 --> 01:26:01.920
Uh-huh.

1444
01:26:01.940 --> 01:26:02.280
Got it.

1445
01:26:02.300 --> 01:26:02.800
Okay.

1446
01:26:04.120 --> 01:26:10.320
Um, so, uh, the, I think the thing that, that we just came to is, um, to what

1447
01:26:10.320 --> 01:26:22.680
extent, so we are over, uh, we are over here to what extent do we need to

1448
01:26:22.680 --> 01:26:28.860
be answering questions about the tools in order to get to a place where we

1449
01:26:28.860 --> 01:26:35.380
understand more or less what we think the work is for, for, for this kind

1450
01:26:35.380 --> 01:26:37.280
of piece of the, of the, of the workflow.

1451
01:26:37.500 --> 01:26:40.320
Are there, are there unknowns here that, that relate to the tools?

1452
01:26:40.320 --> 01:26:41.820
Are there interdependencies there?

1453
01:26:43.900 --> 01:26:47.620
I mean, to me, it's becoming like much clearer now that the order, like the

1454
01:26:47.620 --> 01:26:53.760
dependency was like kind of look at the expected prompts, the case agents will

1455
01:26:53.760 --> 01:27:01.120
have, and then these will inform what tool, tool calls will need to answer those

1456
01:27:01.120 --> 01:27:01.540
questions.

1457
01:27:01.540 --> 01:27:06.940
And then the tool calls will inform how we're going to vectorize the uploaded

1458
01:27:06.940 --> 01:27:07.440
files.

1459
01:27:08.000 --> 01:27:08.840
Uh-huh.

1460
01:27:09.880 --> 01:27:10.640
Uh-huh.

1461
01:27:11.920 --> 01:27:12.200
Yeah.

1462
01:27:12.200 --> 01:27:16.560
Starting with like that, like what question is the prompt going to solve

1463
01:27:16.580 --> 01:27:18.540
and then kind of work backwards into that.

1464
01:27:20.100 --> 01:27:24.860
So, um, so I'm just gonna, I'm not gonna insist on this.

1465
01:27:24.860 --> 01:27:29.380
Is this was just some stuff that's in like a way of looking at what's in a way of,

1466
01:27:29.460 --> 01:27:31.060
you know, what's, what was in my head at the time.

1467
01:27:31.060 --> 01:27:36.660
But, um, what I heard was, um, look at the expected prompts.

1468
01:27:38.340 --> 01:27:42.140
That's going to help us understand what tools we'll need.

1469
01:27:43.500 --> 01:27:43.780
Yep.

1470
01:27:44.440 --> 01:27:51.160
That's going to help us understand how to make choices when vectorizing.

1471
01:27:52.320 --> 01:27:52.520
Yep.

1472
01:27:55.320 --> 01:27:55.920
Um,

1473
01:28:03.200 --> 01:28:05.520
well, it's really, okay.

1474
01:28:06.040 --> 01:28:08.240
Um, good.

1475
01:28:08.280 --> 01:28:08.840
Okay.

1476
01:28:08.980 --> 01:28:13.500
Um, are these, I mean, these feel like, um, things that you can

1477
01:28:13.500 --> 01:28:17.220
clarify, um, that aren't mysterious.

1478
01:28:17.260 --> 01:28:17.780
Yeah.

1479
01:28:19.180 --> 01:28:19.460
Yeah.

1480
01:28:19.460 --> 01:28:20.060
I feel equipped.

1481
01:28:20.100 --> 01:28:20.380
Yeah.

1482
01:28:21.580 --> 01:28:23.620
Um, uh,

1483
01:28:27.940 --> 01:28:28.300
okay.

1484
01:28:28.300 --> 01:28:30.220
Would you take each prompt?

1485
01:28:30.420 --> 01:28:35.240
Like, would this be a good time to take, like, almost treat the prompt like a UI

1486
01:28:35.300 --> 01:28:38.540
element or, or, you know, like I'm going to hit this button and then stuff's going

1487
01:28:38.540 --> 01:28:43.900
to happen, um, like basically breadboard each of the prompts, would that be

1488
01:28:44.500 --> 01:28:48.260
worthwhile exercise or is that unrealistic?

1489
01:28:49.020 --> 01:28:54.540
Um, uh, so we always have the freedom to decide how much detail to go into here.

1490
01:28:54.580 --> 01:28:55.060
Right.

1491
01:28:55.500 --> 01:29:03.260
And, um, and that's going to depend on how much it feels like, uh, it's going to be

1492
01:29:03.280 --> 01:29:09.560
fine or how much we have a little bit of this, like back of our, back of our heads

1493
01:29:09.640 --> 01:29:13.040
doubt, like maybe there's something we're missing there that we should be thinking

1494
01:29:13.040 --> 01:29:13.840
about, you know?

1495
01:29:14.240 --> 01:29:15.880
Um, so it can go either way.

1496
01:29:16.100 --> 01:29:22.840
Um, in terms of like, um, method of like how to map it out.

1497
01:29:22.880 --> 01:29:23.400
Totally.

1498
01:29:24.000 --> 01:29:24.480
Totally.

1499
01:29:24.640 --> 01:29:26.800
Because that's actually how a tool call works.

1500
01:29:26.800 --> 01:29:28.700
As far as I understand, do you know what I mean?

1501
01:29:28.700 --> 01:29:33.520
Like there's going to be the general anatomy of a tool call is there's going

1502
01:29:33.520 --> 01:29:41.760
to be some kind of a prompt and then there is a, uh, uh, um, I mean, of course,

1503
01:29:41.760 --> 01:29:46.320
like, it's not like a, just like an if, if, if, but it kind of feels like that.

1504
01:29:46.360 --> 01:29:47.000
Do you know what I mean?

1505
01:29:47.040 --> 01:29:53.840
Like the, the, the, the LLM is going to, is going to choose from among the

1506
01:29:53.840 --> 01:29:58.200
available tools, like what it thinks it needs to, to use, you know?

1507
01:29:58.200 --> 01:29:59.940
So this is a little bit like, um.

1508
01:30:00.000 --> 01:30:06.320
this is like a black like from so like from submit prompt in or not even what we don't

1509
01:30:06.320 --> 01:30:11.120
call it a prompt but like from like well let's yeah let's just say like we submit the chat message

1510
01:30:16.880 --> 01:30:23.280
there's going to be some sense of um somehow it's going to it's going to

1511
01:30:24.240 --> 01:30:31.040
it's going to go to open ai you know what i mean uh it open open ai is going to choose

1512
01:30:31.040 --> 01:30:35.040
from the available tools and then we're going to get to this place where we have like

1513
01:30:35.600 --> 01:30:48.240
if claims list is invoked uh right um uh and so on you know and and then and and you could totally

1514
01:30:48.240 --> 01:30:55.520
um view this as a ui thing because it's almost like you've navigated to the claims list

1515
01:30:57.360 --> 01:31:01.840
you know what i mean like it's a it's a it's a thing that has a view because there's a there's

1516
01:31:01.840 --> 01:31:07.440
a kind of a claims list has a view model or a read model like there's certain data that it expects to

1517
01:31:07.440 --> 01:31:13.120
get right so if you want to think about it you could do that where you could say like um claims

1518
01:31:13.120 --> 01:31:20.800
list as the tool as as as the tool view and i'm just going to uh

1519
01:31:24.880 --> 01:31:28.240
let's say that i don't remember what we were using you guys had some color for

1520
01:31:28.240 --> 01:31:36.960
uh orange you were using um for uh for a place for a ui place right then that claims list is

1521
01:31:36.960 --> 01:31:44.880
going to have affordances in it i think in this case the tool is just the uh the um we can

1522
01:31:44.880 --> 01:31:49.920
literally just write here like bullet list of claims but if these are going to be linked for

1523
01:31:49.920 --> 01:31:53.680
example and you're going to click on them to jump to a claim that's the kind of thing that you would

1524
01:31:53.680 --> 01:32:00.080
specify as a ui element here right like are these just is this just a list of claims in in plain

1525
01:32:00.080 --> 01:32:05.600
text and it's like claim numbers or are these actually like clickable links that take you to

1526
01:32:05.600 --> 01:32:11.200
the detail pages for those claims uh that's the kind of thing that you are designing here right

1527
01:32:11.200 --> 01:32:17.200
so like what is in this and then um this claims list tool just like everything else we've talked

1528
01:32:17.200 --> 01:32:22.880
about it's going to expect a certain kind of data shape if we think of it from the view standpoint

1529
01:32:22.880 --> 01:32:28.640
here you know what i mean so like if this if this is the view in the tool that openai is going to

1530
01:32:28.640 --> 01:32:36.400
render then it needs to have um well you know i'm the simplest way to say that is that there's

1531
01:32:36.400 --> 01:32:40.320
going to be some list of claims that it's going to take in as the as the input that it needs to

1532
01:32:40.320 --> 01:32:45.280
know and this is giving you the now we we have a way to talk about the shape of the data that

1533
01:32:45.280 --> 01:32:56.880
we need for this tool to be rendered yeah um okay that was just to kind of like unpack the answer to

1534
01:32:56.880 --> 01:33:04.880
to to bruce to your to your to your question there um how do you guys feel um uh do you want

1535
01:33:04.880 --> 01:33:11.440
to um we're an hour and a half in here out of our uh you know three hours max do you feel like you

1536
01:33:11.440 --> 01:33:16.560
need five minutes to breathe you want to take a five minute break and then come back so let's take

1537
01:33:16.560 --> 01:33:23.040
five and uh um uh and uh and take a little breather and then and then come back and then we can decide

1538
01:33:23.040 --> 01:33:27.680
where you want to go next cool okay see you in a bit

1539
01:33:32.480 --> 01:33:34.800
hey bruce you know i can talk i don't know

1540
01:33:35.680 --> 01:33:40.080
no it's good i'm going to grab some some water real quick i'll be right

1541
01:33:53.040 --> 01:33:53.540
you

1542
01:34:23.040 --> 01:34:24.500
you

1543
01:34:53.040 --> 01:34:54.500
you

1544
01:35:00.000 --> 01:35:02.060
you

1545
01:35:30.000 --> 01:35:32.060
you

1546
01:36:00.000 --> 01:36:02.060
you

1547
01:36:30.000 --> 01:36:32.060
you

1548
01:37:00.000 --> 01:37:02.060
you

1549
01:37:30.000 --> 01:37:32.060
you

1550
01:38:00.000 --> 01:38:02.060
you

1551
01:38:30.000 --> 01:38:32.060
you

1552
01:39:00.000 --> 01:39:02.060
you

1553
01:39:30.000 --> 01:39:32.060
you

1554
01:40:00.000 --> 01:40:02.060
you

1555
01:40:30.000 --> 01:40:32.060
you

1556
01:41:00.000 --> 01:41:02.060
you

1557
01:41:30.000 --> 01:41:32.060
you

1558
01:42:00.000 --> 01:42:14.000
all right guys cool so um let's do a little check-in on um kind of where did

1559
01:42:14.000 --> 01:42:18.400
where did we get to versus where you started

1560
01:42:20.720 --> 01:42:23.360
i'll go last

1561
01:42:23.360 --> 01:42:37.040
i guess i'll go first well uh feeling way less worried about this uh project um i see that

1562
01:42:38.560 --> 01:42:39.920
the value of like

1563
01:42:42.320 --> 01:42:48.480
and also the level of definition we should have gone on the breadboarding the current way

1564
01:42:48.480 --> 01:42:58.480
um when you say you you feel less worried what like kind of like why

1565
01:43:01.360 --> 01:43:10.400
what's different maybe it's like your calm uh demeanor that like but um the project therapy

1566
01:43:10.400 --> 01:43:17.840
from ryan yeah no no no demeanor can compete with uh with a problem you don't know how to solve you

1567
01:43:17.840 --> 01:43:29.680
know so i think there's more to it than that um the dependency like um list really clarifies it

1568
01:43:31.360 --> 01:43:38.080
um i mean there's like fewer things to look at it just had like too many things to look at uh-huh

1569
01:43:39.440 --> 01:43:45.200
uh-huh so when so so so when you say there's fewer things to look at kind of what's in your head

1570
01:43:46.080 --> 01:43:51.360
uh now versus before like like like before it was kind of like all the stuff in the project

1571
01:43:51.360 --> 01:43:53.360
was kind of just all there like side by side

1572
01:43:57.280 --> 01:44:04.080
yes there's also um there was like one moment that clarified something i'll get back to like the

1573
01:44:05.200 --> 01:44:12.080
general question was like when you put in the like black box of like vector magic is like what goes

1574
01:44:12.560 --> 01:44:20.080
in the input and the output of that kind of guides what you're going to research and i'm just like a

1575
01:44:20.080 --> 01:44:25.360
general generally curious person so i kind of went off on like what our vector database in general

1576
01:44:25.920 --> 01:44:30.960
so that like reduces it a lot that's a good one okay uh-huh so bounding

1577
01:44:31.600 --> 01:44:36.960
bounding the problem with like in this specific case i need to get from here to here

1578
01:44:37.760 --> 01:44:46.720
uh shrank the research problem yeah yeah and i think a bigger piece is like just that whole like

1579
01:44:46.720 --> 01:44:51.760
dependency graph of like these prompts are expected and we're going to focus on these

1580
01:44:51.760 --> 01:44:57.120
ones and that informs what tool calls we're going to do and the tool calls what we do

1581
01:44:57.120 --> 01:44:59.760
is going to inform like what i'm going to research and how i'm going to do the

1582
01:45:00.000 --> 01:45:02.440
the vectorizing and like just then like

1583
01:45:02.920 --> 01:45:06.080
Piece alone, like if I use that as my north star

1584
01:45:06.080 --> 01:45:09.920
I feel like come back and just like slash everything that doesn't matter to this

1585
01:45:10.560 --> 01:45:12.560
Got it. Uh-huh

1586
01:45:12.840 --> 01:45:14.840
so you have a

1587
01:45:14.840 --> 01:45:16.840
it's kind of like you have like a

1588
01:45:17.040 --> 01:45:20.660
Almost like a plan of attack of like how I'm gonna make my way through this problem

1589
01:45:21.440 --> 01:45:23.440
Yeah, got it. Uh-huh

1590
01:45:23.840 --> 01:45:25.840
Cool, okay

1591
01:45:26.360 --> 01:45:28.360
I

1592
01:45:28.520 --> 01:45:32.700
Anybody else any any any to any difference from where we are now versus where we started

1593
01:45:39.600 --> 01:45:43.040
Yeah, my mind so, you know, I'm um

1594
01:45:44.360 --> 01:45:47.640
I'm not working on this specific project. So I

1595
01:45:48.440 --> 01:45:51.120
Just in this is my first company working

1596
01:45:51.720 --> 01:45:56.960
This is my first time working at a company where shape-up theory was the primary approach so

1597
01:45:58.080 --> 01:46:00.080
I think I'm just getting

1598
01:46:00.760 --> 01:46:03.320
Just more knowledge into shape-up mentality

1599
01:46:04.840 --> 01:46:07.640
Some mental notes on like how to

1600
01:46:09.840 --> 01:46:15.840
Kind of approach breadboarding which I can see now Bruce who you're saying how it felt like we weren't

1601
01:46:16.440 --> 01:46:20.000
Doing like the full unrolled breadboard, you know

1602
01:46:21.480 --> 01:46:23.480
Yeah

1603
01:46:23.560 --> 01:46:24.800
That's cool

1604
01:46:24.800 --> 01:46:29.840
Yeah, so it's nice. Um, just kind of giving that context so I can like have these notes for what I am

1605
01:46:30.240 --> 01:46:32.320
You know doing my own breadboarding in the future

1606
01:46:38.080 --> 01:46:40.040
I think for me

1607
01:46:40.040 --> 01:46:44.640
The last session maybe the second session we had with you, you know, it's kind of a bad connection

1608
01:46:44.640 --> 01:46:46.600
I wasn't quite getting

1609
01:46:46.640 --> 01:46:51.440
The nine grid and just kind of how we got there and Lucien helped me as we were talking through earlier

1610
01:46:51.440 --> 01:46:56.840
but when you were I go back to that skeleton of how you're trying to like slice the problem up because I really

1611
01:46:57.040 --> 01:47:01.920
Didn't know where to start on some of this because like Lucien said there's a lot of stuff right and it's very yeah

1612
01:47:02.720 --> 01:47:06.720
Well, let's do the whole anatomy of that and then the anatomy it's like no we don't do that

1613
01:47:06.720 --> 01:47:10.440
Like what am I touching right now? And I think you know Bruce when your comments was I think we're

1614
01:47:11.520 --> 01:47:15.480
Looking at stuff. We don't necessarily need to right now just for chatting with files, which is true

1615
01:47:15.760 --> 01:47:21.160
So as interesting as some of these, you know strings are to pull on we need a really, you know

1616
01:47:21.680 --> 01:47:26.240
Iron clad focus on this but knowing how to solve the problems along with that

1617
01:47:26.240 --> 01:47:31.440
What Lucien was saying about the left and right bounds of you know, constraining your uncertainty if you will

1618
01:47:32.120 --> 01:47:35.560
Was very helpful to me because I'm sort of someone who's very visual

1619
01:47:35.560 --> 01:47:36.600
I need

1620
01:47:36.600 --> 01:47:40.800
The figmas of the world of whiteboards to kind of sketch things out and see where they come from

1621
01:47:40.800 --> 01:47:44.400
But some of these are like, where do I start right? I think having that

1622
01:47:45.480 --> 01:47:50.680
I know you called it that the the diagramming for the the programmers thing. I need to write that down, but

1623
01:47:52.120 --> 01:47:57.000
Having that approach really gives you a place to start to really get a finger

1624
01:47:57.000 --> 01:47:58.720
Hold on not even a foothold with a finger

1625
01:47:58.720 --> 01:48:03.520
Hold on and then you start kind of clearing that fog of war if you will, right? Yeah

1626
01:48:04.800 --> 01:48:06.800
Super kind of light bulb moment for me

1627
01:48:07.800 --> 01:48:09.800
And

1628
01:48:12.600 --> 01:48:16.840
I'm gonna just intercept Bruce before you go because I wanted to push back on something

1629
01:48:16.840 --> 01:48:22.120
maybe we'll just like put a bookmark on that and talk about that later, but I kind of disagree about like

1630
01:48:22.640 --> 01:48:24.120
You put a message in the chat

1631
01:48:24.120 --> 01:48:29.960
Like we're out of scope of these the entire project is limited to chatting with files only and I questioned that because I think

1632
01:48:30.320 --> 01:48:32.640
that's not a good framing because

1633
01:48:33.160 --> 01:48:39.080
Part of why we named the project like this is they have their custom GPT and it's tedious to like download files from the

1634
01:48:40.680 --> 01:48:42.760
App and then upload to the custom GPT

1635
01:48:43.600 --> 01:48:45.600
but I

1636
01:48:45.640 --> 01:48:51.400
Think the framing it as we're trying to answer like some questions. It's like way more useful

1637
01:48:51.400 --> 01:48:55.640
I guess in like delivering something that's usable because like just chatting with files is like

1638
01:48:55.640 --> 01:49:02.800
It's cool like it removes a step but like what's the concrete value of that? So yeah, I just questioned that

1639
01:49:05.320 --> 01:49:07.320
I mean

1640
01:49:07.320 --> 01:49:09.080
Yeah

1641
01:49:09.080 --> 01:49:11.880
I would say sorry Bruce. I didn't mean to step on your

1642
01:49:12.440 --> 01:49:16.280
Response, but at the end of the day, these are all components to serve a bigger thing

1643
01:49:16.280 --> 01:49:19.720
The the end game is this right now. I'm a case manager

1644
01:49:19.720 --> 01:49:25.560
It's really painful to get a complete coherent picture of where all my clients are think of I have 30 people

1645
01:49:25.560 --> 01:49:28.200
I need to take care of right if I it takes me

1646
01:49:28.200 --> 01:49:34.320
I don't know even 15 minutes to get a complete picture of where this one client is that all this DBQ needs to fill this

1647
01:49:34.320 --> 01:49:39.120
SDR is missing this form is missing. I can't I got a schedule the medical appointment times 30, right?

1648
01:49:39.120 --> 01:49:42.640
That's an enormous amount of cognitive load that an enormous amount of pain

1649
01:49:42.640 --> 01:49:48.000
So this little little section of chatting with files is really important

1650
01:49:48.040 --> 01:49:53.640
Little section of channel files and then that extended AI about form-filling the files

1651
01:49:54.000 --> 01:49:59.600
Then adding in the UI improvements. All of these are three legs of a stool that overall

1652
01:50:00.000 --> 01:50:03.680
completely help have like this perfect assistant

1653
01:50:03.680 --> 01:50:05.420
with perfect memory, right?

1654
01:50:05.420 --> 01:50:07.800
Helping the case managers to serve more people

1655
01:50:07.800 --> 01:50:09.200
and get those claims in faster

1656
01:50:09.200 --> 01:50:11.660
because the government does not move quickly, right?

1657
01:50:11.660 --> 01:50:14.480
But if we can get the claims in, we start the clock running

1658
01:50:14.480 --> 01:50:15.720
and that's the benefit.

1659
01:50:15.720 --> 01:50:20.040
And I think that I am more of a process person.

1660
01:50:20.040 --> 01:50:22.400
So I naturally gravitate toward,

1661
01:50:22.400 --> 01:50:25.400
hey, this process is really, really painful.

1662
01:50:25.400 --> 01:50:28.420
I wanna work on the sexy thing of improving the workflow.

1663
01:50:28.460 --> 01:50:30.300
But to serve that, right?

1664
01:50:30.300 --> 01:50:34.380
The underlying support is chatting with files

1665
01:50:34.380 --> 01:50:35.940
and then form-filling the files.

1666
01:50:35.940 --> 01:50:39.020
So, you know, it's hard for me to kind of like

1667
01:50:39.020 --> 01:50:41.220
pull out of the operational kind of workflow thing

1668
01:50:41.220 --> 01:50:43.340
and just focus on the files thing just for me.

1669
01:50:43.340 --> 01:50:45.060
I know that about myself.

1670
01:50:45.060 --> 01:50:48.140
But I guess I'm trying to, in my role, Ryan,

1671
01:50:48.140 --> 01:50:50.720
I'm trying to provide more, not color commentary,

1672
01:50:50.720 --> 01:50:53.580
but exact, like I'm a champion for this me, right?

1673
01:50:53.580 --> 01:50:55.020
Because I work with them very closely.

1674
01:50:55.020 --> 01:50:56.900
I can speak for them in what they want

1675
01:50:56.900 --> 01:50:57.860
and how they're feeling it

1676
01:50:57.860 --> 01:51:00.600
and how this should kind of terminate,

1677
01:51:00.600 --> 01:51:01.580
the process should terminate.

1678
01:51:01.580 --> 01:51:04.060
So again, solving some of these problems are like,

1679
01:51:04.060 --> 01:51:05.660
it's not like what color is the button,

1680
01:51:05.660 --> 01:51:07.860
but it's more like, hey, they're doing it this way now.

1681
01:51:07.860 --> 01:51:10.220
Can we kill five steps and make it one step?

1682
01:51:10.220 --> 01:51:11.860
You know, so.

1683
01:51:11.860 --> 01:51:16.200
This is a juicy moment in a shaping session

1684
01:51:16.200 --> 01:51:21.040
where the question is coming up of actually what's in scope

1685
01:51:21.040 --> 01:51:23.380
and what's out of scope for this project.

1686
01:51:23.420 --> 01:51:26.020
And this is,

1687
01:51:30.620 --> 01:51:33.240
this is where, okay.

1688
01:51:35.300 --> 01:51:37.540
The way that I like to work on this

1689
01:51:37.540 --> 01:51:41.380
is to use this terminology

1690
01:51:41.380 --> 01:51:43.180
of the shaping versus the framing.

1691
01:51:43.180 --> 01:51:45.140
The shaping is like,

1692
01:51:45.140 --> 01:51:49.500
what are we actually doing in this time box?

1693
01:51:49.500 --> 01:51:50.860
So like this thing we're working on

1694
01:51:50.860 --> 01:51:52.420
that's gonna ship in, what is it?

1695
01:51:52.420 --> 01:51:54.220
Two weeks now from now or whatever.

1696
01:51:55.460 --> 01:51:58.380
This is the thing that we're shaping.

1697
01:51:58.380 --> 01:52:01.500
And then like, but if it's not doing

1698
01:52:01.500 --> 01:52:03.660
what it needs to be doing in order for the customer

1699
01:52:03.660 --> 01:52:07.500
and to actually feel like we gave them the right thing

1700
01:52:07.500 --> 01:52:08.540
and we're making progress,

1701
01:52:08.540 --> 01:52:11.260
then like, well, that's the question, right?

1702
01:52:11.260 --> 01:52:13.420
Like, is this thing actually doing what it needs to do

1703
01:52:13.420 --> 01:52:15.780
in order for it to feel like it's a win?

1704
01:52:15.780 --> 01:52:19.220
And so that's kind of what the framing piece is,

1705
01:52:19.220 --> 01:52:21.940
is like, what are the expectations?

1706
01:52:22.460 --> 01:52:23.780
Like, what's the thing we're actually solving here

1707
01:52:23.780 --> 01:52:24.860
and what are the expectations

1708
01:52:24.860 --> 01:52:27.060
from the sort of stakeholder point of view?

1709
01:52:28.300 --> 01:52:32.260
The thing that is tricky here is,

1710
01:52:33.820 --> 01:52:36.060
Chris, when you mentioned,

1711
01:52:36.060 --> 01:52:39.500
there's this bigger vision that we're trying to get to.

1712
01:52:39.500 --> 01:52:43.060
The thing that kind of feels a little bit fuzzy to me

1713
01:52:43.060 --> 01:52:44.420
as an outsider right now,

1714
01:52:44.420 --> 01:52:47.420
and this is often a challenge in a bigger effort,

1715
01:52:47.420 --> 01:52:52.420
is like, there's the framing at the level of like,

1716
01:52:52.660 --> 01:52:53.860
if you think of this as like,

1717
01:52:53.860 --> 01:52:55.420
this isn't one four-week effort,

1718
01:52:55.420 --> 01:52:58.220
but there's gonna be like many, many slices of work

1719
01:52:58.220 --> 01:52:59.260
to this thing,

1720
01:52:59.260 --> 01:53:02.300
then there's kind of this like monster project

1721
01:53:02.300 --> 01:53:04.060
where like at the end of the monster project,

1722
01:53:04.060 --> 01:53:06.460
they get this value and this is why we're doing it.

1723
01:53:06.460 --> 01:53:09.460
And every step we get closer to that is a win.

1724
01:53:10.820 --> 01:53:15.820
But from the standpoint of selling the weeks,

1725
01:53:16.620 --> 01:53:19.060
like, so kind of more from Bruce, where you sit,

1726
01:53:20.060 --> 01:53:25.060
there is this other dynamic of what is the rate

1727
01:53:25.860 --> 01:53:27.420
of the delivery of the progress?

1728
01:53:27.420 --> 01:53:29.980
Like, how do we make it feel like this four weeks

1729
01:53:29.980 --> 01:53:32.860
we got somewhere where they feel like they're seeing

1730
01:53:32.860 --> 01:53:35.660
the right movement toward that bigger thing?

1731
01:53:35.660 --> 01:53:36.620
Do you know what I mean?

1732
01:53:36.620 --> 01:53:40.300
So like, that to me kind of feels like the thing

1733
01:53:40.300 --> 01:53:42.180
that is fuzzy right now,

1734
01:53:42.180 --> 01:53:44.140
and maybe it's clear and it's just fuzzy to me

1735
01:53:44.140 --> 01:53:45.340
as an outsider,

1736
01:53:45.340 --> 01:53:47.780
which is like when the conversation is happening

1737
01:53:47.780 --> 01:53:50.340
about like, is this in scope or out of scope?

1738
01:53:50.340 --> 01:53:51.900
Is that a conversation about like,

1739
01:53:51.900 --> 01:53:52.940
well, if it's out of scope,

1740
01:53:52.940 --> 01:53:55.220
then we'll just get it in the next round?

1741
01:53:55.220 --> 01:53:58.220
Or is it like, if we leave it out,

1742
01:53:58.220 --> 01:54:00.380
we're gonna go back to them at the end of this four weeks

1743
01:54:00.380 --> 01:54:03.300
and they're gonna say, I'm not impressed.

1744
01:54:03.300 --> 01:54:04.140
Do you know what I mean?

1745
01:54:04.140 --> 01:54:05.660
That's not the expectation that you set

1746
01:54:05.660 --> 01:54:06.980
with what was gonna come out of this.

1747
01:54:06.980 --> 01:54:09.700
So what I would,

1748
01:54:09.700 --> 01:54:14.700
so one way to try and dig into this a little bit

1749
01:54:15.540 --> 01:54:16.740
is to flip it to Bruce.

1750
01:54:16.740 --> 01:54:20.700
And like, Bruce, like from the sales side of this,

1751
01:54:20.700 --> 01:54:21.540
you know what I mean?

1752
01:54:21.540 --> 01:54:25.220
Like, how would you describe like the we're here

1753
01:54:25.220 --> 01:54:27.020
and this is where we're gonna be at the end of this

1754
01:54:27.020 --> 01:54:28.860
in terms of like what you're solving

1755
01:54:28.860 --> 01:54:30.900
or the progress they're expecting to see?

1756
01:54:32.660 --> 01:54:33.500
Great, yeah.

1757
01:54:33.500 --> 01:54:36.140
So this has been a great session.

1758
01:54:36.140 --> 01:54:41.140
So Bruce, I think Ryan just saved this project

1759
01:54:41.340 --> 01:54:43.940
potentially in this last two hours.

1760
01:54:43.940 --> 01:54:46.340
I think this was a dumpster fire

1761
01:54:46.340 --> 01:54:47.660
heading off the end of a cliff.

1762
01:54:47.660 --> 01:54:50.980
So I didn't realize it was quite as bad as it was.

1763
01:54:50.980 --> 01:54:52.300
I got some hints of this

1764
01:54:52.300 --> 01:54:53.860
from some of the comments you guys made yesterday.

1765
01:54:53.860 --> 01:54:56.380
And then we said, let's prioritize, let's get clarity.

1766
01:54:56.380 --> 01:54:59.300
So I think the cool thing is that we're learning like

1767
01:54:59.300 --> 01:55:00.140
getting clear.

1768
01:55:00.000 --> 01:55:04.600
about any project is going to be our secret weapon for everything.

1769
01:55:04.600 --> 01:55:05.840
It's like, oh, this feels hard.

1770
01:55:05.840 --> 01:55:07.160
We feel overwhelmed.

1771
01:55:07.160 --> 01:55:10.760
It probably always comes down to a clarity problem.

1772
01:55:10.760 --> 01:55:12.800
Like, oh, we're not clear.

1773
01:55:12.800 --> 01:55:17.360
And so we were clear.

1774
01:55:17.360 --> 01:55:19.120
And so things got super fuzzy.

1775
01:55:19.120 --> 01:55:21.640
And I think we're coming back to clarity.

1776
01:55:21.640 --> 01:55:24.080
So I think, I mean, I'm going to try

1777
01:55:24.080 --> 01:55:28.440
to separate deeper team process and things just

1778
01:55:28.440 --> 01:55:32.120
to try to bring it just in the context of this project.

1779
01:55:32.120 --> 01:55:33.720
So the first problem I saw was just

1780
01:55:33.720 --> 01:55:37.360
that we have, obviously, right off the bat, scope creep.

1781
01:55:37.360 --> 01:55:41.000
Like, there's a screenshot of the framed document

1782
01:55:41.000 --> 01:55:42.520
we sent to them, which is like, we're

1783
01:55:42.520 --> 01:55:44.480
going to let you chat with 20 different files.

1784
01:55:44.480 --> 01:55:46.320
And you can ask questions about these files.

1785
01:55:46.320 --> 01:55:47.860
And it's going to give you responses.

1786
01:55:47.860 --> 01:55:49.880
Very, very simple.

1787
01:55:49.880 --> 01:55:51.560
We're not going out.

1788
01:55:51.560 --> 01:55:53.280
And we spent a bit of time talking

1789
01:55:53.280 --> 01:55:55.560
about how transcripts work today.

1790
01:55:55.560 --> 01:55:57.680
They don't affect this project, actually.

1791
01:55:57.680 --> 01:55:59.840
And so I had to come back to this framed document.

1792
01:55:59.840 --> 01:56:01.960
Now, there's probably interconnectivities and things

1793
01:56:01.960 --> 01:56:04.280
that we missed on this initial version.

1794
01:56:04.280 --> 01:56:07.000
But we also have to come back from a sales standpoint.

1795
01:56:07.000 --> 01:56:08.600
It's like, hey, we sold this.

1796
01:56:08.600 --> 01:56:09.880
And so initially, it's like, yeah,

1797
01:56:09.880 --> 01:56:11.340
wouldn't it be great if it did all?

1798
01:56:11.340 --> 01:56:13.600
That's exactly what every client is going to want,

1799
01:56:13.600 --> 01:56:14.720
no matter what.

1800
01:56:14.720 --> 01:56:16.680
So we have to be like, yep, but remember files?

1801
01:56:16.680 --> 01:56:18.220
Like, that should have been something

1802
01:56:18.220 --> 01:56:20.920
we pushed back really hard on on the very first call.

1803
01:56:20.920 --> 01:56:23.400
So as soon as they kind of step out of bounds,

1804
01:56:23.400 --> 01:56:25.600
we can remind them that, like, hey, we're.

1805
01:56:25.600 --> 01:56:29.480
So there's kind of a skill of managing the client's

1806
01:56:29.480 --> 01:56:31.760
expectations now.

1807
01:56:31.760 --> 01:56:34.640
So luckily, we have a call with them in a few hours.

1808
01:56:34.640 --> 01:56:37.480
So we'll get ready for that and prep all of this

1809
01:56:37.480 --> 01:56:40.440
to make more sense for that.

1810
01:56:40.440 --> 01:56:42.160
So I think that was my first thing, which

1811
01:56:42.160 --> 01:56:43.880
is like, yeah, scope creep.

1812
01:56:43.880 --> 01:56:47.220
Now, whether it should have been files only or not,

1813
01:56:47.220 --> 01:56:49.320
I think this is like a different conversation.

1814
01:56:49.320 --> 01:56:50.960
Maybe we wrote down this, but now it's

1815
01:56:50.960 --> 01:56:52.480
like other conversations have led to,

1816
01:56:52.480 --> 01:56:54.160
oh, what we really need is a this.

1817
01:56:54.160 --> 01:56:57.440
Well, that thing could be two six-week projects.

1818
01:56:57.440 --> 01:56:58.160
I don't know.

1819
01:56:58.160 --> 01:57:02.080
Like, we've just blown right past any kind of framing

1820
01:57:02.080 --> 01:57:03.760
documents that we put together.

1821
01:57:03.760 --> 01:57:05.680
And we kind of treat the framed document

1822
01:57:05.680 --> 01:57:06.880
kind of like our agreement.

1823
01:57:06.880 --> 01:57:10.360
We have an overarching legal agreement with them,

1824
01:57:10.360 --> 01:57:12.640
just on all code and how IP works and all that.

1825
01:57:12.640 --> 01:57:15.200
Right, but that's like the master of services agreement.

1826
01:57:15.200 --> 01:57:17.120
And then the framed document is more

1827
01:57:17.120 --> 01:57:20.520
like the scope of work for an individual chunk, right?

1828
01:57:20.520 --> 01:57:22.720
Yeah, and it needs to be clear.

1829
01:57:22.720 --> 01:57:24.120
And we do have some out of bounds.

1830
01:57:24.640 --> 01:57:25.680
I'm pulling it up now.

1831
01:57:25.680 --> 01:57:28.800
To me, it still feels pretty clear where it was.

1832
01:57:28.800 --> 01:57:32.960
But it's as if we're talking about two different projects,

1833
01:57:32.960 --> 01:57:37.480
from what we're talking about even in the breadboarding here.

1834
01:57:37.480 --> 01:57:39.720
Because I think that framed document

1835
01:57:39.720 --> 01:57:43.040
is there to give the client boundaries

1836
01:57:43.040 --> 01:57:44.400
and to give our team boundaries.

1837
01:57:44.400 --> 01:57:45.880
Because I think that's where it does.

1838
01:57:45.880 --> 01:57:47.560
You start to open the code, and you start to say, well,

1839
01:57:47.560 --> 01:57:48.280
this connects to that.

1840
01:57:48.280 --> 01:57:49.720
And then, well, really, what we need is a that.

1841
01:57:49.720 --> 01:57:51.720
And then it's like, all of a sudden, there's 50 projects.

1842
01:57:51.720 --> 01:57:53.680
Like, there's just so many ways we can go.

1843
01:57:53.720 --> 01:57:55.720
And there's not this clear direction.

1844
01:57:55.720 --> 01:58:01.800
So I think that's something I think we got way off on.

1845
01:58:01.800 --> 01:58:03.720
And I think just giving ourselves enough time.

1846
01:58:03.720 --> 01:58:05.960
So maybe when we framed it, maybe we

1847
01:58:05.960 --> 01:58:09.680
didn't frame it all the way we should.

1848
01:58:09.680 --> 01:58:14.400
The document we produced and came out that we have today

1849
01:58:14.400 --> 01:58:16.520
isn't accurate.

1850
01:58:16.520 --> 01:58:18.200
I don't know if we have the flexibility

1851
01:58:18.200 --> 01:58:20.480
to change framed documents.

1852
01:58:20.480 --> 01:58:22.120
That would be a question for you, Ryan,

1853
01:58:22.120 --> 01:58:25.840
like your thoughts on, can framed documents change,

1854
01:58:25.840 --> 01:58:26.840
or should they?

1855
01:58:29.640 --> 01:58:31.200
We've done other framed projects.

1856
01:58:31.200 --> 01:58:35.480
I did one with Aira for an e-commerce ERP platform.

1857
01:58:35.480 --> 01:58:38.040
And I feel like that was very tight.

1858
01:58:38.040 --> 01:58:40.160
It felt very concrete to me, I think to Aira,

1859
01:58:40.160 --> 01:58:41.000
and even the client.

1860
01:58:41.000 --> 01:58:46.040
He could see very specifics what this is and isn't going to be.

1861
01:58:47.000 --> 01:58:55.680
So I think we have done it, in other cases, a bit better.

1862
01:58:55.680 --> 01:58:59.840
And again, this project was started maybe four weeks ago.

1863
01:58:59.840 --> 01:59:01.440
The one I did with Aira was two weeks.

1864
01:59:01.440 --> 01:59:04.920
So we're probably just a couple of weeks ahead in our thinking.

1865
01:59:04.920 --> 01:59:07.880
Maybe an extra call from you or a few extra breadboarding

1866
01:59:07.880 --> 01:59:09.960
sessions better at it.

1867
01:59:09.960 --> 01:59:12.760
So it's literally like we're just doing it on,

1868
01:59:12.760 --> 01:59:14.600
like we're kind of learning it and also

1869
01:59:14.600 --> 01:59:18.720
trying to implement all the things we're learning at the same time.

1870
01:59:18.720 --> 01:59:22.600
So I'm not picking at anything.

1871
01:59:22.600 --> 01:59:25.560
If anything, I think the biggest culprit on this whole project

1872
01:59:25.560 --> 01:59:28.840
was me, because we started it, framed it,

1873
01:59:28.840 --> 01:59:30.440
and then I handed it off to you guys.

1874
01:59:30.440 --> 01:59:32.840
And then I sort of backed off for like a week or two.

1875
01:59:32.840 --> 01:59:35.040
And so this is my first re-entry into the project.

1876
01:59:35.040 --> 01:59:38.240
So I think that was the biggest sort of bug in the system,

1877
01:59:38.240 --> 01:59:43.000
was me taking a step back too quickly.

1878
01:59:43.080 --> 01:59:46.480
And then we also threw in another developer.

1879
01:59:46.480 --> 01:59:51.280
So I think any of my feedback is not Lucian or Chris,

1880
01:59:51.280 --> 01:59:54.280
or I feel like you guys are trying, like we all kind of get

1881
01:59:54.280 --> 01:59:57.000
the vision of where the client wants, but we have to kind of put it

1882
01:59:57.000 --> 01:59:59.680
through our process, or else we're going to end up with this.

1883
01:59:59.680 --> 02:00:01.240
Because what's.

1884
02:00:00.000 --> 02:00:01.760
This happens all the time.

1885
02:00:01.800 --> 02:00:04.240
We're like, Oh, we're going to add, let's just add these other two things.

1886
02:00:04.280 --> 02:00:06.640
And now like we missed the four week deadline.

1887
02:00:06.640 --> 02:00:09.960
Now we're in week six, we're burning hours faster.

1888
02:00:09.960 --> 02:00:12.040
Like we're over budget and we're over schedule.

1889
02:00:12.040 --> 02:00:17.320
And the clients like now has these crazy expectations because I'd rather like do

1890
02:00:17.320 --> 02:00:20.720
something very small and throw one or like, and do that really, really well.

1891
02:00:21.000 --> 02:00:22.880
Then now they kind of have this expectations.

1892
02:00:22.880 --> 02:00:26.720
Like I'm going to, like, as soon as we saw things, like, I want to ask it, um,

1893
02:00:26.880 --> 02:00:30.560
what, you know, some of these prompts we should just be like, Oh, that, Nope.

1894
02:00:30.600 --> 02:00:31.240
Can't do that one.

1895
02:00:31.240 --> 02:00:31.880
Can't do that one.

1896
02:00:31.920 --> 02:00:33.240
Oh, we'll do this one that you sent us.

1897
02:00:33.240 --> 02:00:33.880
We can do that.

1898
02:00:33.900 --> 02:00:37.680
You know, like just immediate feedback to them, like, so that they're, because

1899
02:00:37.680 --> 02:00:40.360
we can see, Oh, like they think they're getting something they think they're

1900
02:00:40.360 --> 02:00:43.840
getting, like, you know, we're doing a and B, but they think they're getting

1901
02:00:43.880 --> 02:00:48.320
like CD and F like, we need to just be very clear about, Hey, remember this

1902
02:00:48.320 --> 02:00:52.320
document now, and that's where a lot of this, this fuzziness comes from.

1903
02:00:52.800 --> 02:00:53.480
Um, I think so.

1904
02:00:53.720 --> 02:00:57.960
I just want to make sure that we, we hold that frame document as like, okay.

1905
02:00:57.960 --> 02:01:00.120
Like at the beginning of every meeting is like, let's read

1906
02:01:00.120 --> 02:01:01.360
through this document again.

1907
02:01:01.400 --> 02:01:02.480
Like, are we on track?

1908
02:01:02.880 --> 02:01:05.320
Are we like within the boundaries of this?

1909
02:01:05.640 --> 02:01:09.240
Because, and again, we, it would also be worth potentially looking at that

1910
02:01:09.480 --> 02:01:14.840
document and saying like, well, this, like we probably framed it 40% and it

1911
02:01:14.840 --> 02:01:16.640
could have been 80 or 90%, right?

1912
02:01:16.640 --> 02:01:18.720
Like it maybe isn't very clear yet.

1913
02:01:19.160 --> 02:01:24.080
Um, and maybe it wasn't concrete enough that when we're in a call with a client

1914
02:01:24.080 --> 02:01:26.480
or we're digging into the code, we can't just be like, Oh no, we've

1915
02:01:26.480 --> 02:01:27.960
identified this issue already.

1916
02:01:28.160 --> 02:01:29.440
And we have a hard boundary here.

1917
02:01:29.440 --> 02:01:30.840
So I want to pause there.

1918
02:01:30.840 --> 02:01:33.520
Cause I had a couple of questions in there for Ryan and I'll

1919
02:01:33.520 --> 02:01:35.360
let you answer mainly around.

1920
02:01:35.400 --> 02:01:37.560
So that's fantastic.

1921
02:01:37.600 --> 02:01:42.560
Um, um, so one thing I just want to highlight, um, uh, is like in the

1922
02:01:42.560 --> 02:01:46.360
contrast between Bruce, the way you answered and Chris, the way you answered.

1923
02:01:46.920 --> 02:01:52.960
I want to make sure that like, um, uh, I, those things often, they come out to

1924
02:01:52.960 --> 02:01:58.120
the team sounding like complete opposite answers, like one of them is a big

1925
02:01:58.120 --> 02:02:02.160
expansive vision and one of them is like, we're not doing this and we're not

1926
02:02:02.160 --> 02:02:05.680
doing that and we're like only scoping it down to this and they can feel very

1927
02:02:05.680 --> 02:02:09.720
like opposed and this can also lead to like, if we don't, if we don't have a

1928
02:02:09.720 --> 02:02:13.560
tool for how to like talk about those things, it can lead to a bunch of weird

1929
02:02:13.600 --> 02:02:16.200
misunderstandings on the, under the surface of things also.

1930
02:02:16.520 --> 02:02:21.800
And like those things, they totally, um, they totally coexist.

1931
02:02:21.880 --> 02:02:24.600
There's always the bigger vision that we want to get to.

1932
02:02:25.040 --> 02:02:32.880
And, and like the, the, the specific sale of a certain amount of money in a certain

1933
02:02:32.880 --> 02:02:35.120
amount of time to make a certain amount of progress.

1934
02:02:35.640 --> 02:02:42.120
So, um, I like to think of the frame as the increment of progress.

1935
02:02:43.720 --> 02:02:51.400
So like the frame is the agreement that we made for a specific unit of like time,

1936
02:02:51.400 --> 02:02:54.120
money, team, you know what I mean?

1937
02:02:54.600 --> 02:03:00.320
And there is a thing that's bigger than the frame that it, especially when you're

1938
02:03:00.320 --> 02:03:04.920
the one who's in touch with the SME, like that is like the, the reason we're doing

1939
02:03:04.920 --> 02:03:07.000
all of this, there is this like bigger thing.

1940
02:03:07.400 --> 02:03:11.920
And the word I like to use for that, just to put a word on it is strategy.

1941
02:03:12.200 --> 02:03:14.280
Like there's a bigger thing we're trying to do.

1942
02:03:14.440 --> 02:03:20.760
There's a reason why we're investing in this particular increment of progress,

1943
02:03:20.960 --> 02:03:29.760
you know, and, um, when, when we're, when we're doing the framing that a lot of

1944
02:03:29.760 --> 02:03:37.000
the, the, the coming to the agreement of the frame with the client is about talking

1945
02:03:37.000 --> 02:03:41.880
through that bigger thing and getting, getting aligned on what the

1946
02:03:41.880 --> 02:03:43.520
meaningful smaller step is.

1947
02:03:45.600 --> 02:03:46.040
Do you know what I mean?

1948
02:03:46.040 --> 02:03:50.480
So like, like how does this relate to the bigger picture is very

1949
02:03:50.480 --> 02:03:53.880
much, um, part of the sales process.

1950
02:03:55.440 --> 02:03:55.940
Yep.

1951
02:03:56.200 --> 02:04:02.040
And then, and then when, once we're in shaping, um, we don't want to have this

1952
02:04:02.040 --> 02:04:05.960
bad, we can easily get into this frustration of like, yeah, but there's

1953
02:04:05.960 --> 02:04:07.600
all this other stuff that we're not doing.

1954
02:04:08.240 --> 02:04:12.520
And so we kind of want to feel like the frame is like our, it's like our

1955
02:04:12.520 --> 02:04:15.560
umbrella, you know, that like, it's like our little safety zone.

1956
02:04:15.880 --> 02:04:22.920
Like, like if I'm in the frame, I'm fine because like when we're in the shaping

1957
02:04:22.920 --> 02:04:26.360
work, we're already downstream from the, you know what I mean?

1958
02:04:26.360 --> 02:04:29.960
Like the project's been bought, like the, the, the thing's been signed,

1959
02:04:29.960 --> 02:04:31.720
like the expectations have been set.

1960
02:04:32.120 --> 02:04:36.040
So everything that we do inside of the shaping should fee, we, we, we want

1961
02:04:36.040 --> 02:04:41.000
this, the frame to feel like these really solid outer walls, you know

1962
02:04:41.000 --> 02:04:41.440
what I mean?

1963
02:04:41.440 --> 02:04:42.800
Like they're the outer walls.

1964
02:04:42.800 --> 02:04:47.320
We're like, as long as I'm inside of these walls and I'm, and I'm, I'm

1965
02:04:47.320 --> 02:04:49.240
meeting, it's actually the same thing.

1966
02:04:49.280 --> 02:04:52.840
Um, Lucian that happened with the black box on the vector thing.

1967
02:04:53.480 --> 02:04:56.080
When we put the inputs and outputs on it, you're like, oh, okay.

1968
02:04:56.120 --> 02:05:00.280
I just have to make this box work in this specific situation.

1969
02:05:00.000 --> 02:05:13.000
That's exactly what a frame is. It's like, oh, we're in this moment in time and we've got these inputs and we need to produce this output and like this is where we're trying to get to together and then everything is all about the inside of the box.

1970
02:05:13.000 --> 02:05:19.000
So that's kind of like where those two things like fit together.

1971
02:05:20.000 --> 02:05:35.000
I'll pause before. I have some other things based, Bruce, what you were saying about framing, especially I want to comment on changing the frame and the feeling about is the frame detailed enough, but I just wanted to see all that monologue.

1972
02:05:35.000 --> 02:05:55.000
Is there anything, did that, did that make any sense? Like, yeah, like, like, like, like, like Chris, does that connect at all? Or, or, like, is there a, is there a link there between the bigger picture and the smaller picture? Or, I don't know if I was too abstract or?

1973
02:05:55.000 --> 02:06:16.000
No, I think that was helpful. I mean, obviously, when you said, do, do we think there, we can see that increment of progress? I mean, certainly I do. And actually Lucian, we were talking about this the other day of the SME I'm talking with, the team member, the client, he already uses his own chat GBT.

1974
02:06:17.000 --> 02:06:43.000
So one of my little tasks, and he kind of sent some stuff over during our conversation was, what does, what do they currently do with their chat GBT? Right, exactly. How do they use it in the workflow? And what outputs is it giving them? Because that's the baseline, right? The, my worst nightmare is, oh, hey, we spent four weeks and built this thing. And it was like the same capabilities or slightly less than what they have. Like, that's the worst thing, right? It's like, what did we spend money on? What do we buy?

1975
02:06:43.000 --> 02:07:01.000
You know, so to Bruce's kind of the sales side of it, I want to know what that, what the current state is. And we want to be this, because we promised this, right, whatever this is, right. And then even better, if, you know, we have more time, we do this plus one, right, which is out of scope. But, you know, that's kind of how I look at it.

1976
02:07:01.000 --> 02:07:30.000
So the increment of progress to me is obviously all the stuff in that document that Bruce pasted in the zoom chat. But the way that it works has to work with, I mean, the way they work, that sounds like it's a tautology, but right. So, so when we framed it, when we shaped it, when we gave them this, this list of things, obviously, we listened to them, we got the input, and we, we trusted our earlier selves to put this together intelligently, right.

1977
02:07:30.000 --> 02:07:54.000
And so I definitely want to make sure we don't diverge from this and give them what we promised, right. So I'm trying to fill that in by getting this, all this contextual information. So to do it right. On the other hand, I will say, I think, I don't think we wasted time in the sense of understanding this, because this is the first time we've done anything with, I've never heard of a vector database before this project, right, or a rag, right.

1978
02:07:54.000 --> 02:08:23.000
So I think there is some useful learning, both from an intellectual curiosity perspective, but also to inform how we move forward. I think role, to Bruce's point, role definition is important. And knowing who's doing what, when is of course important, right. If you're on the kitchen, and you're all, you know, cutting potatoes, when someone's supposed to be boiling potatoes, that's not going to make the meal go faster. Right. So there are a lot of little learnings that we've, we've, we've equipped ourselves with. And I think, again, to Bruce's point, we're getting smarter and sharper.

1979
02:08:24.000 --> 02:08:48.000
Two weeks ago selves are going to be smarter than four weeks ago of ourselves. And we just have to keep, you know, improving how we do this. And again, I do agree with Bruce that if we keep this as our sort of our North Star, and we don't deviate from this, like that'll help us a lot more as we say, Hey, is this in scope or out scope and be very, very stringent with ourselves and with the client about what the expectations are. So yeah, this is this is eye opening.

1980
02:08:49.000 --> 02:08:56.000
Um, cool. Okay. Bruce, to your question about like the detailing of the frame and changing the frame.

1981
02:08:58.000 --> 02:08:59.000
The

1982
02:09:00.000 --> 02:09:22.000
I'll make a couple comments. And, you know, framing is just like shaping in the sense that like, if we really want to be like, how would we shape this differently than we have a shaping session, and then we experience it, you know what I mean? So the framing is like something where it's its own work step where we can also dive in there, you know.

1983
02:09:23.000 --> 02:09:49.000
And the, the, the, the two kind of places where we, how to say this, the more that the frame describes an outcome, and the less that it describes exactly how we're going to do it, the more that we are going to feel secure that it's clear.

1984
02:09:50.000 --> 02:09:53.000
So like, if the frame is like,

1985
02:09:54.000 --> 02:09:59.000
I'm going to be able to get answers to XYZ via chat instead of

1986
02:10:00.000 --> 02:10:04.920
of like in terms of like almost like in terms of expected prompts, you know what I mean?

1987
02:10:05.120 --> 02:10:09.520
Like, I'm going to be able to get out of all the different things that take time.

1988
02:10:11.200 --> 02:10:14.600
Let me actually, I'll just share my screen here.

1989
02:10:17.160 --> 02:10:26.080
This problem definition here is, as I understand it, the problem definition of the bigger thing,

1990
02:10:26.680 --> 02:10:28.000
not the four weeks.

1991
02:10:30.840 --> 02:10:31.640
Is that right?

1992
02:10:33.640 --> 02:10:38.800
I mean, in my mind, it was the works, we're really speaking a lot to like large sets of PDF

1993
02:10:38.800 --> 02:10:42.800
documents. And then we're speaking very much to that, like, we don't mention the word

1994
02:10:42.800 --> 02:10:47.720
transcripts here, where I think that to me, the strategy or that bigger picture thing is, I

1995
02:10:47.720 --> 02:10:54.360
want to get a holistic view of everything that's that's going on with this, you know, with

1996
02:10:54.360 --> 02:11:01.240
this person somehow, it could be, it was kind of, it could could be both, I do see the word

1997
02:11:01.240 --> 02:11:06.000
transcript now, like in the second paragraph. So that could just been like, a slip of

1998
02:11:06.000 --> 02:11:11.560
again, maybe us not being super clear on the distinction between transcripts and files, for

1999
02:11:11.560 --> 02:11:12.240
example.

2000
02:11:18.600 --> 02:11:22.880
I'm trying to see what we can do here without fully going into a framing session.

2001
02:11:24.920 --> 02:11:25.720
Mm hmm.

2002
02:11:25.760 --> 02:11:31.320
The feedback's good. Like if that sounds to you more like the bigger vision of like, here's what

2003
02:11:31.320 --> 02:11:36.680
we want to do. The next three to four months, like would be this this overarching, you know,

2004
02:11:37.040 --> 02:11:38.120
place we're gonna land.

2005
02:11:39.600 --> 02:11:44.960
Yeah, so what I heard when you described the problem to me, I even I did a quick sketch,

2006
02:11:44.960 --> 02:11:50.040
because I didn't know if we might need it. I did a quick sketch of like, we're doing 150 claims a

2007
02:11:50.040 --> 02:11:55.280
month right now. Each claim involves 20 plus files, interviews and lots of data. Currently,

2008
02:11:55.280 --> 02:11:58.920
they're pulling files into the computer, dropping some things into chat GPT, trying to

2009
02:11:58.920 --> 02:12:07.440
summarize. And like, this workflow is like, it's too time consuming for this day. And it's

2010
02:12:07.440 --> 02:12:11.480
bottlenecking their ability to take on the amount of cases that they need to take on, they

2011
02:12:11.480 --> 02:12:15.800
want to be able to take on more cases. Is this kind of like what the bigger picture is about?

2012
02:12:16.560 --> 02:12:24.440
Yes. So yeah, that's sort of big pictures. Like, yeah, making it more efficient, make it more

2013
02:12:24.440 --> 02:12:31.640
accurate. And right now, there's only like one or two people that can do this, like perfect, like

2014
02:12:31.680 --> 02:12:35.760
really well. And that's because we can get through this company. Yeah, because they're doing it all

2015
02:12:35.760 --> 02:12:39.040
just out of experience, not with with any tools.

2016
02:12:39.040 --> 02:12:46.280
Right. So there's a story here, there's a basic, the basic framing here. And this is, so this,

2017
02:12:46.280 --> 02:12:56.200
this is like, this is the crash course in framing. The thing that starts to really help is to split

2018
02:12:56.200 --> 02:13:02.360
the world into two things. Only describing the current reality and what's bad about that. And

2019
02:13:02.840 --> 02:13:09.320
then describing the after state. And like, this is like an after that I want to pay for, you know, I

2020
02:13:09.320 --> 02:13:14.840
don't actually understand the after state here well enough yet, to to to really nail it down.

2021
02:13:14.840 --> 02:13:20.520
There's something where, for some reason, only one or two people being able to do this is not okay

2022
02:13:20.520 --> 02:13:26.120
with them to the point that they're actually willing to spend money on improving this. I don't

2023
02:13:26.120 --> 02:13:29.880
know if that's like, I don't know what the story is behind that. Do you know what I mean? So I don't

2024
02:13:29.880 --> 02:13:36.520
know if this what what what what it means. Sometimes you can look at this and you can see, obviously,

2025
02:13:36.520 --> 02:13:42.280
this sounds inefficient. But there's a difference between it's inefficient. And we're going to live

2026
02:13:42.280 --> 02:13:48.200
with it versus it's inefficient. And now is the time to invest. Yeah. So from a sales standpoint,

2027
02:13:48.200 --> 02:13:53.160
I would, I would want to, you know what I mean, like, I don't want to, you know, I don't want to

2028
02:13:53.720 --> 02:13:57.560
I would want to, you know what I mean, like, have that answered.

2029
02:14:01.080 --> 02:14:05.800
Then that if that's the bigger thing, I don't actually know what this is, like,

2030
02:14:10.200 --> 02:14:15.320
that what I would be trying to do, then is if if if this is the bigger thing, and we can't,

2031
02:14:15.320 --> 02:14:21.720
we can't achieve, like, in four weeks, I don't think the promise you're making is you will now

2032
02:14:21.720 --> 02:14:25.480
be able to scale this because other people in the org will be able to do the workflow.

2033
02:14:26.200 --> 02:14:26.760
Yeah, definitely.

2034
02:14:26.760 --> 02:14:33.320
So then the so then the question is, how do we define a segment of this? How do we define a

2035
02:14:33.320 --> 02:14:43.400
slice of this? But in the same way? So like, the thing the thing that that I'm that that I'm

2036
02:14:43.400 --> 02:14:49.240
missing. And by the way, framing is a bit of a black art, to be honest, it's it's a blacker art

2037
02:14:49.240 --> 02:14:55.960
than shaping is. It's so like, we hadn't even heard of it until we started working with you.

2038
02:14:56.760 --> 02:14:59.800
Yeah, none of so none of this is criticism.

2039
02:15:00.000 --> 02:15:02.960
because it's really like a dark art, you know.

2040
02:15:03.900 --> 02:15:07.100
What very often happens is the problem

2041
02:15:07.100 --> 02:15:08.360
is easier to talk about,

2042
02:15:08.360 --> 02:15:11.060
so we do a pretty good job of describing the problem,

2043
02:15:12.600 --> 02:15:14.740
but the outcome is harder to talk about.

2044
02:15:14.740 --> 02:15:17.600
This is how you will know that we succeeded.

2045
02:15:18.520 --> 02:15:20.500
And so what we tend to do,

2046
02:15:20.500 --> 02:15:23.840
because the outcome is hard to talk about,

2047
02:15:23.840 --> 02:15:26.480
we tend to describe the outcome

2048
02:15:26.480 --> 02:15:29.280
as a bunch of supply-side stuff that's gonna get built.

2049
02:15:30.200 --> 02:15:31.280
So at the end of this,

2050
02:15:31.280 --> 02:15:33.160
you're gonna have a Ragnology engine,

2051
02:15:33.160 --> 02:15:35.240
and you're gonna have a chat interface,

2052
02:15:35.240 --> 02:15:37.360
and you know what I mean?

2053
02:15:37.360 --> 02:15:40.400
There's a short list of success just at the bottom,

2054
02:15:40.400 --> 02:15:43.240
because I'd make sure we didn't miss it down here.

2055
02:15:43.240 --> 02:15:45.400
Like we had the success criteria,

2056
02:15:45.400 --> 02:15:47.160
so these are the checkboxes to hit.

2057
02:15:54.480 --> 02:15:59.480
So it says answers based only on client PDFs.

2058
02:15:59.560 --> 02:16:01.160
Like it was very specific.

2059
02:16:01.160 --> 02:16:02.880
This is great.

2060
02:16:02.880 --> 02:16:05.960
So here, I'm gonna hack a frame, okay?

2061
02:16:07.880 --> 02:16:10.440
We're not gonna care about some ideal thing.

2062
02:16:10.440 --> 02:16:13.120
I'm just gonna hack this. So here's what I'm gonna do.

2063
02:16:13.120 --> 02:16:14.720
I'm gonna ...

2064
02:16:16.880 --> 02:16:18.480
We're just gonna do like this.

2065
02:16:20.120 --> 02:16:21.640
You can do it in Miro if you want.

2066
02:16:21.640 --> 02:16:23.600
We can translate it.

2067
02:16:23.600 --> 02:16:28.080
I actually like having it in the shape as a reference,

2068
02:16:28.080 --> 02:16:30.240
so I'll just ...

2069
02:16:30.240 --> 02:16:32.480
I have a boiled-down version of it,

2070
02:16:32.480 --> 02:16:35.760
even a way of kind of sticking it in the corner.

2071
02:16:35.760 --> 02:16:37.360
Sorry, one second.

2072
02:16:40.120 --> 02:16:42.559
Guys, can you give me 30 seconds just to receive a key

2073
02:16:42.559 --> 02:16:43.920
that someone is dropping off?

2074
02:16:43.920 --> 02:16:45.760
That's great. I'll be right back.

2075
02:16:49.040 --> 02:16:52.400
Sweet. I think that's good.

2076
02:16:55.440 --> 02:16:56.520
So I think ...

2077
02:16:56.559 --> 02:16:58.559
I mean, one thing I'm seeing as well,

2078
02:16:59.879 --> 02:17:03.120
that success criteria is a north star.

2079
02:17:03.120 --> 02:17:05.080
The more I read the frame documents,

2080
02:17:05.080 --> 02:17:06.840
I'm more confident than ever

2081
02:17:06.840 --> 02:17:09.280
that we did a great job at actually framing it.

2082
02:17:09.280 --> 02:17:11.600
I just think we sort of glanced at it

2083
02:17:11.600 --> 02:17:13.200
and then forgot about it

2084
02:17:13.200 --> 02:17:17.080
and then didn't really treat it like the legal agreement

2085
02:17:17.080 --> 02:17:17.879
of what we were ...

2086
02:17:17.879 --> 02:17:19.480
That's exactly what happened.

2087
02:17:20.480 --> 02:17:22.879
Yeah.

2088
02:17:23.000 --> 02:17:24.760
That's nice, but it's cooler.

2089
02:17:30.200 --> 02:17:32.000
So that's important, because I think ...

2090
02:17:32.000 --> 02:17:33.520
I was thinking we were like ...

2091
02:17:33.520 --> 02:17:35.320
It seems confusing, but I'm glad ...

2092
02:17:35.320 --> 02:17:37.280
So I'm glad we were ...

2093
02:17:37.280 --> 02:17:38.840
That's actually one of the pros of this,

2094
02:17:38.840 --> 02:17:40.520
is that I'm seeing like, hey,

2095
02:17:40.520 --> 02:17:43.600
we did actually do a pretty good job two weeks ago

2096
02:17:43.600 --> 02:17:45.200
when we presented it to them.

2097
02:17:46.360 --> 02:17:48.879
I do think there's probably some coordination

2098
02:17:48.879 --> 02:17:51.600
we need to make after this call.

2099
02:17:53.440 --> 02:17:55.360
Because we started on the second.

2100
02:17:56.639 --> 02:17:58.400
There's little things, like the dates

2101
02:17:58.400 --> 02:18:00.000
weren't even in the project in Notion.

2102
02:18:00.000 --> 02:18:02.440
I'm not being picky, but I just added them now.

2103
02:18:02.440 --> 02:18:04.240
So we have 19 days left.

2104
02:18:05.879 --> 02:18:08.680
I've been texting with MJ on the side as well.

2105
02:18:08.680 --> 02:18:10.559
So I'm going to do a call with him.

2106
02:18:10.559 --> 02:18:14.000
I think we got off a little bit on a bumpy start there.

2107
02:18:14.000 --> 02:18:17.400
And then right now, it's like we've planned to do June 30th.

2108
02:18:17.400 --> 02:18:20.520
So what we do need to do is if we get super clear today

2109
02:18:20.559 --> 02:18:23.840
or tomorrow, either push it out a week potentially,

2110
02:18:23.840 --> 02:18:26.680
because that is one of the knobs we can turn

2111
02:18:26.680 --> 02:18:27.440
to give you guys more free.

2112
02:18:27.440 --> 02:18:28.879
But let's coordinate that after.

2113
02:18:28.879 --> 02:18:31.959
Ryan doesn't have to be dragged into that, those details.

2114
02:18:33.240 --> 02:18:35.160
Okay, so we left off.

2115
02:18:35.160 --> 02:18:37.840
I was just going to hack a frame together.

2116
02:18:39.280 --> 02:18:45.160
So I'm just going to take this problem definition,

2117
02:18:47.559 --> 02:18:49.360
and I'm going to skip the solution.

2118
02:18:50.280 --> 02:18:53.000
And I'm just going to go straight to success criteria.

2119
02:18:56.799 --> 02:19:01.799
And then I'm going to add one thing that's missing.

2120
02:19:14.440 --> 02:19:15.600
What this is not.

2121
02:19:16.600 --> 02:19:19.240
There's a, yeah, we didn't do a lot of this.

2122
02:19:22.680 --> 02:19:25.480
Like we did a little out of bounds section,

2123
02:19:25.480 --> 02:19:29.320
but it's just two things which would have been,

2124
02:19:29.320 --> 02:19:31.160
we did not mention transcripts specifically,

2125
02:19:31.160 --> 02:19:33.080
although it's not mentioned anywhere in the solution

2126
02:19:33.080 --> 02:19:34.719
or the success criteria.

2127
02:19:34.719 --> 02:19:39.719
So the way that I'm thinking when I'm framing

2128
02:19:39.920 --> 02:19:44.920
is all framing conversations

2129
02:19:46.879 --> 02:19:50.120
are very important conversations

2130
02:19:50.120 --> 02:19:52.640
where misunderstandings are deeply costly.

2131
02:19:55.960 --> 02:20:00.960
This is, if I'm going to sign a contract with a client,

2132
02:20:00.000 --> 02:20:01.000
You know what I mean?

2133
02:20:01.000 --> 02:20:08.340
Like we're kicking off a project and what I need is for it to be so boiled down that

2134
02:20:08.340 --> 02:20:10.120
no one will miss anything.

2135
02:20:10.120 --> 02:20:13.200
You know what I mean?

2136
02:20:13.200 --> 02:20:18.280
If I really, really boil it down, I can be sure that the things that I'm worried about

2137
02:20:18.280 --> 02:20:24.880
as a, to sort of protect my own skin, you know, are not going to get missed.

2138
02:20:24.880 --> 02:20:34.000
So this does a really good job of explaining that you understand the problem that you're

2139
02:20:34.000 --> 02:20:41.760
solving, that you are aligned on like why this matters and why this should happen now.

2140
02:20:41.760 --> 02:20:48.300
This is a perfectly, perfectly acceptable way to describe like, this is how you know

2141
02:20:48.300 --> 02:20:53.680
that the project worked, you know.

2142
02:20:53.680 --> 02:21:05.160
The more that you leave out the how, the gladder you will be because that's your latitude.

2143
02:21:05.160 --> 02:21:08.500
That's your freedom of movement, do you know what I mean?

2144
02:21:08.500 --> 02:21:13.020
And the more that you, the opposite is also true.

2145
02:21:13.020 --> 02:21:17.140
The more that you put in, like there's going to be a vector DB there or there's going to

2146
02:21:17.140 --> 02:21:23.100
be, because maybe you don't even, there's a, like, there is, I mean, even the state

2147
02:21:23.100 --> 02:21:25.500
of the art of rag is moving so much.

2148
02:21:25.500 --> 02:21:29.740
Like you could decide at the last minute, like we actually think we can just get by

2149
02:21:29.740 --> 02:21:32.660
with a bigger context window and blah, blah, blah, you know what I mean?

2150
02:21:32.660 --> 02:21:40.220
Like the more that this is boiled down to this is what you're going to be expecting

2151
02:21:40.220 --> 02:21:47.820
to test when we bring this to you, you know, or this is how you're going to know that we

2152
02:21:47.820 --> 02:21:49.660
did our job.

2153
02:21:49.660 --> 02:21:53.780
And then there's going to be something here that, you know, you kind of don't want to

2154
02:21:53.780 --> 02:21:59.440
get sucked into, or for you was important, that was out of scope, that you kind of like,

2155
02:21:59.440 --> 02:22:04.340
it was in the negotiation, but it kind of needs to be noted when it was signed.

2156
02:22:04.340 --> 02:22:15.860
And let's say that that was like, this is about files, not transcripts.

2157
02:22:15.860 --> 02:22:24.380
If I give this to a client and it's literally only this, and then there's going to be, there's

2158
02:22:24.380 --> 02:22:29.780
going to be, you know, there's going to be, if this is a, I mean, there's going to be

2159
02:22:29.780 --> 02:22:36.740
like a, like a sort of a time cost piece, time cost schedule piece, you know what I mean?

2160
02:22:36.740 --> 02:22:37.340
Appetite, yeah.

2161
02:22:41.500 --> 02:22:47.340
From the, from the framing perspective, you know, of like working with a client, like I

2162
02:22:47.340 --> 02:22:49.740
wouldn't even use the appetite language.

2163
02:22:49.740 --> 02:22:53.060
I would literally just be like, this is a, this is a time commitment.

2164
02:22:53.100 --> 02:22:57.020
These are, this is like, there's a, there's a moment in time where both of us expect to

2165
02:22:57.020 --> 02:23:00.260
meet up again and, and I'm, we're supposed to be delivering something.

2166
02:23:00.300 --> 02:23:00.900
Do you know what I mean?

2167
02:23:01.420 --> 02:23:10.220
And then this, um, uh, appetite is a little bit more of, um, uh, it's a little

2168
02:23:10.220 --> 02:23:13.220
bit more of a product team concept, you know, and it's a little hard to

2169
02:23:13.220 --> 02:23:15.540
translate that, um, to the client.

2170
02:23:15.560 --> 02:23:18.740
You know, it's, it's a good concept to have in mind when you're in the

2171
02:23:18.740 --> 02:23:23.380
negotiation of the frame, but the output of the frame is actually like a harder

2172
02:23:23.380 --> 02:23:26.900
agreement of like, this is just schedule, you know, that's at least how

2173
02:23:27.100 --> 02:23:31.620
I find it helpful to separate those, you know, cause for me, appetite is a

2174
02:23:31.620 --> 02:23:34.420
very, um, design process word.

2175
02:23:34.900 --> 02:23:35.580
You know what I mean?

2176
02:23:35.620 --> 02:23:39.140
Like, like, so when I'm early in a negotiation, trying to frame something,

2177
02:23:39.180 --> 02:23:40.740
I'm trying to feel out your budget.

2178
02:23:40.740 --> 02:23:41.240
Right.

2179
02:23:41.540 --> 02:23:44.060
Like, like how important is this to you?

2180
02:23:44.060 --> 02:23:45.940
Like how, how urgent is it?

2181
02:23:45.980 --> 02:23:49.180
And, and, and I'm starting to suss out appetite based on that.

2182
02:23:49.580 --> 02:23:54.140
And I'm like, so maybe we could do a, maybe we could do like a, let's do

2183
02:23:54.140 --> 02:23:56.660
three, four week projects on this and sign that as one thing.

2184
02:23:56.700 --> 02:23:58.340
And they're like, no, no, no, we're not ready for that.

2185
02:23:58.740 --> 02:23:59.620
And you're like, oh, okay.

2186
02:23:59.660 --> 02:24:00.260
You know what I mean?

2187
02:24:00.260 --> 02:24:03.980
Like, uh, it's, it's more like in the negotiation, but then the kind of

2188
02:24:03.980 --> 02:24:07.380
when we're done, it's, it's kind of like baked, um, anyway, that's how I think

2189
02:24:07.380 --> 02:24:18.380
about it, um, is this, um, uh, uh, the, the, yeah, I don't think we need

2190
02:24:18.380 --> 02:24:21.540
to go further than that, like, do you, let me, let me hand it back to you.

2191
02:24:21.580 --> 02:24:22.060
Yeah.

2192
02:24:22.060 --> 02:24:24.900
Um, um, do you, what do you see in this?

2193
02:24:25.900 --> 02:24:27.260
Like what's helpful here.

2194
02:24:29.740 --> 02:24:33.020
Um, and just to summarize, like you have, you're thinking of having a problem

2195
02:24:33.100 --> 02:24:36.700
section, the success section, like a checklist like this, and then having

2196
02:24:36.700 --> 02:24:39.260
a specific, like not section.

2197
02:24:39.580 --> 02:24:42.100
And then maybe the fourth is just this budget schedule.

2198
02:24:42.580 --> 02:24:43.860
Uh, like, so, yeah.

2199
02:24:43.860 --> 02:24:51.300
So if you're thinking of it as a document, then it's, there's, then

2200
02:24:51.540 --> 02:24:55.860
there's going to be something like, uh, you know, usually there's going to be

2201
02:24:55.860 --> 02:24:56.460
something.

2202
02:25:00.000 --> 02:25:03.000
like fees or cost or price or whatever.

2203
02:25:03.000 --> 02:25:05.080
Yeah, okay.

2204
02:25:05.080 --> 02:25:07.480
Usually this is kind of the package here.

2205
02:25:07.480 --> 02:25:10.520
And how this gets done,

2206
02:25:10.520 --> 02:25:13.480
I mean, that's at your discretion.

2207
02:25:13.480 --> 02:25:16.520
As long as these, the crispier these are,

2208
02:25:16.520 --> 02:25:19.080
and the better you get at spelling these out,

2209
02:25:19.080 --> 02:25:21.560
the more you're gonna feel like,

2210
02:25:21.560 --> 02:25:23.680
I'm just responsible for delivering the outcome.

2211
02:25:23.680 --> 02:25:27.140
And I'm gonna have five different ideas

2212
02:25:27.140 --> 02:25:30.320
and shaping of like how I actually do that, you know?

2213
02:25:30.320 --> 02:25:31.160
Right.

2214
02:25:33.300 --> 02:25:35.000
I think we, yeah, that's good.

2215
02:25:36.700 --> 02:25:40.660
And I think in every meeting, including this one,

2216
02:25:40.660 --> 02:25:42.420
like if Chris or Lucian,

2217
02:25:42.420 --> 02:25:44.740
like we're talking about anybody to the project,

2218
02:25:44.740 --> 02:25:47.460
like coming to this for the context,

2219
02:25:47.460 --> 02:25:51.820
like we probably should have started here today

2220
02:25:51.820 --> 02:25:54.260
because it would have directed what we did.

2221
02:25:54.260 --> 02:25:56.860
We learned a lot as far as like how to shape some things up,

2222
02:25:56.860 --> 02:25:58.940
but all of that was mostly around like transcripts

2223
02:25:58.940 --> 02:26:01.940
and stuff, so we can apply that to the file side.

2224
02:26:01.940 --> 02:26:03.220
But this would have been, you know,

2225
02:26:03.220 --> 02:26:05.300
just even for framing this conversation

2226
02:26:05.300 --> 02:26:09.220
would have been helpful to revisit this as a starting point.

2227
02:26:10.680 --> 02:26:13.860
Because ideally, I guess if this is,

2228
02:26:13.860 --> 02:26:14.780
because if you had seen like,

2229
02:26:14.780 --> 02:26:16.020
oh, this is not about transcripts,

2230
02:26:16.020 --> 02:26:17.980
and then we spent 10 minutes talking about transcripts,

2231
02:26:17.980 --> 02:26:20.060
we were like, wait, didn't we say we're not doing this?

2232
02:26:20.060 --> 02:26:22.580
Like, you know, it'd be so clear and obvious to people,

2233
02:26:22.580 --> 02:26:23.540
like you're saying.

2234
02:26:23.660 --> 02:26:26.500
So I think that's helpful

2235
02:26:26.500 --> 02:26:28.860
if we do spend the time to get it like that.

2236
02:26:28.860 --> 02:26:32.180
I do like the idea of keeping the freedom.

2237
02:26:32.180 --> 02:26:35.860
So actually not having the skipping,

2238
02:26:35.860 --> 02:26:38.780
and this is where we were very fuzzy about shaping,

2239
02:26:38.780 --> 02:26:41.020
like we were doing shaping so badly

2240
02:26:41.020 --> 02:26:43.360
that it looked closer to framing.

2241
02:26:43.360 --> 02:26:46.340
And even at that, it wasn't a very good frame

2242
02:26:46.340 --> 02:26:47.820
to summarize where we were.

2243
02:26:47.820 --> 02:26:49.380
So I was like, it's shaping,

2244
02:26:49.380 --> 02:26:51.860
this is technically our kind of shaped,

2245
02:26:51.860 --> 02:26:54.060
packaged document template.

2246
02:26:54.060 --> 02:26:55.740
I'm gonna, I'm making a new one now,

2247
02:26:55.740 --> 02:26:57.780
it will just be for framing,

2248
02:26:57.780 --> 02:26:59.820
which would be the thing we hand to the client.

2249
02:26:59.820 --> 02:27:01.860
And then everything else we can do,

2250
02:27:01.860 --> 02:27:03.940
you know, in using whatever tools we want,

2251
02:27:03.940 --> 02:27:05.700
because I think that's trying to pick

2252
02:27:05.700 --> 02:27:08.180
like the right level of abstraction,

2253
02:27:08.180 --> 02:27:10.580
like to show, hey, this, like I said,

2254
02:27:10.580 --> 02:27:13.060
general problem, here's success, super clear.

2255
02:27:13.060 --> 02:27:14.420
In case you're worried about this,

2256
02:27:14.420 --> 02:27:16.460
these are the three things we're not doing.

2257
02:27:16.460 --> 02:27:18.900
And then getting them to agree to that.

2258
02:27:18.900 --> 02:27:21.980
Because I think what I'm most concerned with,

2259
02:27:21.980 --> 02:27:23.540
because Chris was like, oh, the worst thing we could do

2260
02:27:23.540 --> 02:27:25.180
is like they spend four weeks of budget,

2261
02:27:25.180 --> 02:27:27.020
and then, you know, they don't get what,

2262
02:27:27.020 --> 02:27:29.540
no, the worst thing we can do is we said we would do A,

2263
02:27:29.540 --> 02:27:32.620
and then we, or A, and then they expect A, B, and C,

2264
02:27:32.620 --> 02:27:34.700
and then we sort of have to deliver A,

2265
02:27:34.700 --> 02:27:36.380
and sort of have to deliver B.

2266
02:27:36.380 --> 02:27:38.220
So like, the best thing we could do is like,

2267
02:27:38.220 --> 02:27:40.620
we said this, we held you to it for four weeks,

2268
02:27:40.620 --> 02:27:41.460
so the next time we,

2269
02:27:41.460 --> 02:27:42.460
because what I would be,

2270
02:27:42.460 --> 02:27:44.700
I would love for them to do the four weeks,

2271
02:27:44.700 --> 02:27:46.020
pay for it, get the,

2272
02:27:46.020 --> 02:27:48.620
deliver exactly what's on the screen now.

2273
02:27:49.340 --> 02:27:50.540
And then when they'd say, oh, but what about X, Y, and Z?

2274
02:27:50.540 --> 02:27:51.740
Oh yeah, remember when we like,

2275
02:27:51.740 --> 02:27:53.420
you remember that we agreed to like this thing?

2276
02:27:53.420 --> 02:27:54.500
Oh, okay, like we held it.

2277
02:27:54.500 --> 02:27:55.580
And they shouldn't be surprised

2278
02:27:55.580 --> 02:27:58.660
because we should be like reminding them of this weekly,

2279
02:27:58.660 --> 02:28:00.540
as we, every conversation we have.

2280
02:28:01.740 --> 02:28:03.540
Keeping this moral star.

2281
02:28:03.540 --> 02:28:05.780
The more that you boil this down,

2282
02:28:07.060 --> 02:28:12.060
the thing I aim for is literally one paragraph

2283
02:28:12.340 --> 02:28:14.940
of the size of this first paragraph for problem,

2284
02:28:14.940 --> 02:28:16.940
and not more than that for solution.

2285
02:28:16.940 --> 02:28:17.780
Do you know what I mean?

2286
02:28:18.780 --> 02:28:22.820
The more that it becomes really, really, really crispy,

2287
02:28:22.820 --> 02:28:26.380
like small, but extremely crispy and clear what it is,

2288
02:28:26.380 --> 02:28:28.820
the easier it will be to repeat it

2289
02:28:28.820 --> 02:28:31.180
every time you ever make a decision.

2290
02:28:31.180 --> 02:28:32.020
You know what I mean?

2291
02:28:32.020 --> 02:28:33.940
Like you want the frame to almost be like

2292
02:28:33.940 --> 02:28:38.420
this little like mantra of like, why did they hire me?

2293
02:28:38.420 --> 02:28:39.260
Do you know what I mean?

2294
02:28:39.260 --> 02:28:40.300
Like, why did they hire me?

2295
02:28:40.300 --> 02:28:44.060
Like, oh, I'm here in these four weeks to do this,

2296
02:28:44.060 --> 02:28:45.540
to solve this thing.

2297
02:28:45.540 --> 02:28:46.380
You know what I mean?

2298
02:28:47.020 --> 02:28:49.180
And you would say the exact same thing to the client

2299
02:28:49.180 --> 02:28:50.300
if they walked into the room

2300
02:28:50.300 --> 02:28:52.500
and they started to talk to you about some other scope,

2301
02:28:52.500 --> 02:28:54.900
you'd say the same thing to each other as engineers,

2302
02:28:54.900 --> 02:28:56.460
or you'd say the same thing internally,

2303
02:28:56.460 --> 02:28:58.420
like between sales and engineering or whatever.

2304
02:28:58.420 --> 02:28:59.260
You know what I mean?

2305
02:28:59.260 --> 02:29:02.900
Like, it's that very clear mission statement

2306
02:29:02.900 --> 02:29:04.620
kind of of what we're doing.

2307
02:29:04.620 --> 02:29:06.100
Can't believe I just said mission statement,

2308
02:29:06.100 --> 02:29:09.300
but you know, mission definition of what the mission is.

2309
02:29:09.420 --> 02:29:10.260
No.

2310
02:29:13.380 --> 02:29:15.820
Yeah, and to Chris's point,

2311
02:29:18.100 --> 02:29:20.220
these should feel,

2312
02:29:21.140 --> 02:29:24.500
I mean, in the negotiation of getting to these,

2313
02:29:24.500 --> 02:29:27.660
it should feel like, yes, this is valuable.

2314
02:29:27.660 --> 02:29:29.540
This is moving them ahead of the baseline

2315
02:29:29.540 --> 02:29:32.260
of what they already have, right?

2316
02:29:32.260 --> 02:29:35.700
Yeah, because I think part of this is also setting up,

2317
02:29:35.700 --> 02:29:37.140
which is not really clear here,

2318
02:29:37.140 --> 02:29:40.260
but it is kind of giving the foundation

2319
02:29:40.260 --> 02:29:41.620
to do a bunch of other things.

2320
02:29:41.620 --> 02:29:44.620
Like we kind of, they don't have anything like this

2321
02:29:44.620 --> 02:29:46.180
in their code base yet.

2322
02:29:46.180 --> 02:29:48.540
So setting this first piece just to do PDFs

2323
02:29:48.540 --> 02:29:51.740
is gonna kind of pave the way for transcripts

2324
02:29:51.740 --> 02:29:53.420
and doing other things and bolting on

2325
02:29:53.420 --> 02:29:55.620
more types of problems we can ask, you know?

2326
02:29:55.620 --> 02:29:57.940
So I think, but keeping it clear,

2327
02:29:57.940 --> 02:30:00.100
because what they don't wanna do is they,

2328
02:30:00.000 --> 02:30:04.960
you know open it on the first test run and start throwing in like random prompts at it

2329
02:30:05.520 --> 02:30:12.240
um and then just yeah like it's sort of quote not working at that point right yeah um

2330
02:30:13.520 --> 02:30:21.600
uh there's a there's a the i'll just share with you briefly the the so this is kind of like um

2331
02:30:22.800 --> 02:30:28.080
uh what a document can look like when we need when we need an agreement

2332
02:30:29.040 --> 02:30:36.720
um when when i'm just kind of framing for myself or or to get to this document or to have a quick

2333
02:30:36.720 --> 02:30:43.040
reference during shaping i use something a little more informal and that's um that's this thing that

2334
02:30:43.040 --> 02:30:49.440
that that we that i had in the mural which is i just have two boxes and one is context and one

2335
02:30:49.440 --> 02:30:55.360
is outcome but we could to use your to use your language we could just say problem and success

2336
02:30:55.360 --> 02:31:05.280
for example it's the same meaning um and uh here's where um i'll very often just i like to

2337
02:31:05.280 --> 02:31:15.040
use just little uh boxes again um and i'll kind of quick and dirty just spell out the things where

2338
02:31:15.280 --> 02:31:23.760
um they uh like the um they have a manual process uh involving chat gpt

2339
02:31:26.160 --> 02:31:35.040
not enough people can do that process um so like as an outcome of this project we are going to

2340
02:31:35.920 --> 02:31:48.880
um we're going to pave the way for um uh the system to offer integrated um like ai um

2341
02:31:49.600 --> 02:31:56.240
uh um it's kind of like um it's kind of like we're gonna we're gonna we're gonna pave the way

2342
02:31:56.800 --> 02:32:03.360
for the system to eventually offer all of the things our experts are our experts are doing

2343
02:32:04.320 --> 02:32:06.320
by hand with chat gpt

2344
02:32:08.400 --> 02:32:12.640
something like that so like the paving the way is part of the outcome do you know what i mean

2345
02:32:12.640 --> 02:32:17.600
so like this can't just be like a hack that like doesn't give you any leverage in the follow-up

2346
02:32:17.600 --> 02:32:24.160
project you know what i mean but then there's more to it which is like um the there's this

2347
02:32:24.160 --> 02:32:30.240
other thing of like um what is the what is it actually going to do at the end of this four weeks

2348
02:32:30.800 --> 02:32:33.280
so there's something here um

2349
02:32:37.360 --> 02:32:39.200
about the client pdfs

2350
02:32:42.880 --> 02:32:44.400
there's something where um

2351
02:32:46.000 --> 02:32:51.840
uh i mean i could paste these in i kind of feel like there's a bigger thing because some of these

2352
02:32:51.840 --> 02:32:58.880
things are more like um just requirements and some of these things are more outcomes like data

2353
02:32:59.760 --> 02:33:06.800
being secure isn't an outcome data being secure is a requirement uh it's not like oh we paid you

2354
02:33:06.800 --> 02:33:11.200
we you worked for four weeks and now our insecure data became secure do you know what i mean like

2355
02:33:12.480 --> 02:33:19.600
so the outcome is more around um being intuitive for example is also like more of a requirement

2356
02:33:20.400 --> 02:33:26.000
um so like there's a there's something that they can do here that they couldn't do before

2357
02:33:26.880 --> 02:33:28.800
they can ask um

2358
02:33:31.040 --> 02:33:35.440
they can ask questions and get context-rich answers about client pdfs i'll just put that in

2359
02:33:35.440 --> 02:33:36.400
as a first draft

2360
02:33:43.360 --> 02:33:51.680
would you spell out um that distinction of of success criteria verse uh like success first the

2361
02:33:52.000 --> 02:33:58.240
the uh or the outcome verse the um the requirements like those are almost the

2362
02:33:58.240 --> 02:34:03.520
clients are kind of like a footnote of like yeah and some of this should be pretty obvious like

2363
02:34:04.080 --> 02:34:10.240
the requirements the data secure like that's you know these are actually your your um fine print

2364
02:34:11.520 --> 02:34:16.320
right you know what i mean they're not really that it's not like what's the thing that you're

2365
02:34:16.320 --> 02:34:21.120
going to test and demo you know what i mean like it's it's it's uh it's more like the fine print

2366
02:34:21.120 --> 02:34:27.760
you know so like there's going to be these other requirements we've been using the green box that

2367
02:34:27.760 --> 02:34:34.560
you had with like the requirements um yeah the fit check fit check yeah that's right yeah that's

2368
02:34:34.560 --> 02:34:39.760
that's yeah that's i mean they they kind of fall under that right um uh but they're not like the

2369
02:34:39.760 --> 02:34:47.920
big thing that this is about you know um so um i'm gonna do this very roughly um um

2370
02:34:51.120 --> 02:34:59.920
i'm just going to mark this as red um

2371
02:35:00.000 --> 02:35:05.880
If I were really trying to understand this better, I would want to understand a little

2372
02:35:05.880 --> 02:35:10.520
bit about what the before world is of case managers not being able to ask questions about

2373
02:35:10.520 --> 02:35:14.960
client PDFs, like what is going wrong with that?

2374
02:35:14.960 --> 02:35:18.900
And I would like to say something about that here, do you know what I mean?

2375
02:35:18.900 --> 02:35:23.720
Because that is a little bit more specific than this and it justifies like why, because

2376
02:35:23.720 --> 02:35:26.600
what's going to happen is they're going to say like, well, this project is just about

2377
02:35:26.600 --> 02:35:32.000
case managers being able to ask client, being able to ask about PDFs and somebody is going

2378
02:35:32.000 --> 02:35:36.000
to say, well, that's not enough, like we want transcripts too.

2379
02:35:36.000 --> 02:35:42.320
And if you can say, yeah, but today this is what's going wrong because case managers can't

2380
02:35:42.320 --> 02:35:49.360
ask questions about PDFs, that's the thing back to like, oh, right, okay, that is a win.

2381
02:35:49.360 --> 02:35:50.360
Yeah.

2382
02:35:50.360 --> 02:35:53.520
Do you know what I mean?

2383
02:35:53.520 --> 02:36:06.000
So the homework I would have here is, you don't happen to have an answer for that offhand,

2384
02:36:06.000 --> 02:36:07.000
do you?

2385
02:36:07.000 --> 02:36:12.640
About like what happens in the current world without this new ability to ask questions

2386
02:36:12.640 --> 02:36:13.640
about files?

2387
02:36:13.640 --> 02:36:14.640
I do.

2388
02:36:14.760 --> 02:36:15.760
I do.

2389
02:36:15.760 --> 02:36:18.640
They spend more time looking for the files themselves.

2390
02:36:18.640 --> 02:36:28.640
So literally, yeah.

2391
02:36:28.640 --> 02:36:32.200
Can you put that into a little bit more of a specific context to...

2392
02:36:32.200 --> 02:36:36.720
Case managers spend more time trying to manage the case, sorry, case managers spend more

2393
02:36:36.720 --> 02:36:40.600
time trying to figure out how to manage the case and actually doing the work of managing

2394
02:36:40.600 --> 02:36:41.680
the case.

2395
02:36:41.680 --> 02:36:45.560
So that's good at a very, very high level.

2396
02:36:45.560 --> 02:36:50.160
And now the challenge is I'm trying to bring it down to this project.

2397
02:36:50.160 --> 02:36:56.440
So case managers spend more time trying to find the files that they need than actually

2398
02:36:56.440 --> 02:37:07.600
doing the work, finding the files takes more time than actually doing the work of the files.

2399
02:37:07.600 --> 02:37:12.080
Now here's a dumb question.

2400
02:37:12.080 --> 02:37:13.080
Why is this bad?

2401
02:37:13.080 --> 02:37:14.080
Like, I get it.

2402
02:37:14.080 --> 02:37:15.380
Like there's everything takes too much time.

2403
02:37:15.380 --> 02:37:19.720
There's a million things that are inefficient in life.

2404
02:37:19.720 --> 02:37:21.600
How is this?

2405
02:37:21.600 --> 02:37:25.800
I'd say it's hard.

2406
02:37:25.800 --> 02:37:29.640
One thing we did identify, and this is where we actually framed up a different project

2407
02:37:29.640 --> 02:37:36.320
and they put this on for later, was that the entire file interface is terrible.

2408
02:37:36.920 --> 02:37:38.400
They're in these weird folders.

2409
02:37:38.400 --> 02:37:40.560
You can't see how many files are.

2410
02:37:40.560 --> 02:37:43.640
You have to click on each one like two or three times just to get it to download or

2411
02:37:43.640 --> 02:37:44.640
to view it.

2412
02:37:44.640 --> 02:37:50.560
So we had like a kind of an updated two-week, we had a two-week project we did for just

2413
02:37:50.560 --> 02:37:52.200
fixing like how to get to files.

2414
02:37:52.200 --> 02:37:56.240
So I think that this problem would probably fit more under that frame.

2415
02:37:56.240 --> 02:38:01.080
And that's where we kind of spliced it here to say, hey, this is this problem over here.

2416
02:38:01.080 --> 02:38:04.440
Like we're going to solve this problem with this outcome.

2417
02:38:04.440 --> 02:38:10.480
Whereas the other project was really, they kind of had the same problem, but like we're

2418
02:38:10.480 --> 02:38:11.480
solving it in different way.

2419
02:38:11.480 --> 02:38:13.360
One, we're using a chat interface.

2420
02:38:13.360 --> 02:38:20.920
The other one is making it just a much better UI to get better context of how I see files.

2421
02:38:20.920 --> 02:38:27.680
So there's another project where of improving the UI for the filing upload and stuff like

2422
02:38:27.680 --> 02:38:28.680
that.

2423
02:38:28.920 --> 02:38:35.040
Yeah, and just visibility of seeing, you know, hey, there's three files here in this section

2424
02:38:35.040 --> 02:38:37.280
versus, you know, there should be 10 or whatever.

2425
02:38:37.280 --> 02:38:38.280
I don't know.

2426
02:38:38.280 --> 02:38:39.280
Got it.

2427
02:38:39.280 --> 02:38:40.280
Okay.

2428
02:38:40.280 --> 02:38:44.680
And we have that shaped in a similar or framed, I'll say, in a similar format.

2429
02:38:44.680 --> 02:38:45.880
Right.

2430
02:38:45.880 --> 02:38:51.000
They just saw the AI shiny object and we went with that.

2431
02:38:51.000 --> 02:38:52.000
Okay.

2432
02:38:52.000 --> 02:38:55.560
So there's another, one of the things I try to do is because we're bouncing back and forth

2433
02:38:55.680 --> 02:39:00.360
also between the levels that Chris introduced and both of these levels are important.

2434
02:39:00.360 --> 02:39:04.000
One of the things that I sometimes find is helpful when I'm trying to frame something

2435
02:39:04.000 --> 02:39:12.120
is Bob calls it, sometimes we think we're selling the sandwich, but we're actually selling

2436
02:39:12.120 --> 02:39:18.520
just like, but we're actually just selling the mayo and like, we're the mayonnaise, not

2437
02:39:18.520 --> 02:39:19.520
the sandwich.

2438
02:39:19.520 --> 02:39:23.680
And the thing is that like this project is like, it's just like mayonnaise.

2439
02:39:23.800 --> 02:39:26.800
There's going to be another project that is lettuce and another project that is tomato.

2440
02:39:26.800 --> 02:39:28.240
You know what I mean?

2441
02:39:28.240 --> 02:39:33.840
And so what I'm doing is like, I, in my head, I just think of it as like the sandwich, but

2442
02:39:33.840 --> 02:39:39.180
like it's, it's, it's, it's like the, the bigger thing that we're trying to get to.

2443
02:39:39.180 --> 02:39:42.400
It's also like, sometimes like what we call like the, it's like the monster, like the

2444
02:39:42.400 --> 02:39:47.080
big thing that all these projects add up to, you know, and sometimes it helps to be able

2445
02:39:47.080 --> 02:39:49.400
to pull things up and be like, oh, okay.

2446
02:39:49.400 --> 02:39:55.240
So like this, this is like a general part of the big problem that like they're spending,

2447
02:39:55.240 --> 02:39:58.920
like, like Chris said, they're spending too much time on like the management of all this

2448
02:39:58.920 --> 02:40:00.080
stuff instead of actually doing it.

2449
02:40:00.000 --> 02:40:06.800
the work. So there's another slice, there's another like Mayo project which is improving

2450
02:40:06.800 --> 02:40:14.640
the file UI. This project, it's like, what I'm struggling with is to see the connection, is it

2451
02:40:14.640 --> 02:40:23.200
that in the current world files are all over the place, it's hard to like find the files that you

2452
02:40:23.200 --> 02:40:30.960
need, they're not all centralized in the system, and so like in order, when they do this manual

2453
02:40:30.960 --> 02:40:38.240
workflow of getting the files together to upload them to, if they want to like ask questions,

2454
02:40:38.240 --> 02:40:42.320
okay let me ask you this, are they already asking questions about files using AI,

2455
02:40:42.320 --> 02:40:48.960
or is this new behavior for them? They are, they're using their own

2456
02:40:48.960 --> 02:40:54.560
chatgpt and chatbots to do that. So they're like uploading the file kind of like one by one and

2457
02:40:54.560 --> 02:41:01.840
then asking questions about it? So I don't know exactly how they're doing that because they do

2458
02:41:01.840 --> 02:41:06.160
have some qualms about you know privacy and all that stuff, that's why I did reach out to this

2459
02:41:06.160 --> 02:41:12.800
me to see, hey can you just record what you're doing? I think what it's, he's, I think

2460
02:41:12.800 --> 02:41:19.280
safe to say, he's using it like an assistant, like hey, so for John Smith, they have these

2461
02:41:19.280 --> 02:41:26.800
five forms, he's supposed to send me this next Tuesday, right, and then he has tons of

2462
02:41:26.800 --> 02:41:32.960
post-it notes, like literally pasted around his monitor, like the battle things. So the reason I

2463
02:41:32.960 --> 02:41:37.440
kind of talked about files and trying to find them is because that's a first step, but then

2464
02:41:37.440 --> 02:41:41.120
there's not just files, like what are they done with the files, right, what's the newest update,

2465
02:41:41.680 --> 02:41:47.360
are there things on the calendar, so to speak, that he said, okay, next Tuesday they're going

2466
02:41:47.360 --> 02:41:52.000
to do this, oh the medical provider is on vacation, they have to come back in three weeks, so I got

2467
02:41:52.000 --> 02:41:56.720
to schedule that. So it's like a lot of to-dos and that's really the pain point, to get a

2468
02:41:56.720 --> 02:42:01.200
bigger picture of what's supposed to happen now, later, and then just, you know, what has happened

2469
02:42:01.200 --> 02:42:08.480
in the past. How does that relate to asking questions about PDFs? Because there's a

2470
02:42:08.480 --> 02:42:14.800
status to each PDF, is it filled out yet, has it been approved by the medical provider, have they

2471
02:42:14.800 --> 02:42:19.520
been submitted, it's kind of like phase gates, right, so each document on its own has to go

2472
02:42:19.520 --> 02:42:27.600
through these different statuses to be approved, and are there any, and this is

2473
02:42:28.400 --> 02:42:36.560
now, not later, are there any things in the PDF that require other interventions, so for instance,

2474
02:42:36.880 --> 02:42:44.000
do they have a claim that could increase, have, like for instance, maybe a veteran has

2475
02:42:44.000 --> 02:42:48.640
done their own work and submitted something before, right, that might be an increase, so now

2476
02:42:48.640 --> 02:42:54.000
they're a client of VR, and so VR is looking at, well, we can bump that percentage from 11% to

2477
02:42:54.000 --> 02:43:01.120
18%, you know, if the max is 20, or something like that. Got it, yeah, okay, so I'm gonna

2478
02:43:01.120 --> 02:43:06.480
jump a little bit because we're running out of time, and I'm gonna do a not great job of

2479
02:43:06.480 --> 02:43:16.800
this, but I'm gonna just pull out a couple things. So, like, I'm just calling out two

2480
02:43:16.800 --> 02:43:21.840
things I heard you say that were really meaningfully concrete, you know, like, especially good was, like,

2481
02:43:22.720 --> 02:43:26.400
there's something about, like, status that they have to, like, look at the files one by one to, like,

2482
02:43:26.400 --> 02:43:30.480
figure out the status based on what's in there, and then there's also this question about, like, are

2483
02:43:30.480 --> 02:43:34.640
there, there's things they need to know, like, about interventions and so on that require, kind of,

2484
02:43:34.640 --> 02:43:40.720
like, taking a lot of time through the files, and if they can just ask questions about these files,

2485
02:43:41.440 --> 02:43:46.160
this is going to be a significant time savings and productivity for them. This is an example

2486
02:43:46.880 --> 02:43:52.880
of, like, there's this bigger problem, right, of, like, we can't scale our workflow,

2487
02:43:53.440 --> 02:43:57.440
and then, like, if we could just ask questions about PDFs,

2488
02:43:59.040 --> 02:44:03.120
you know, this problem, this, these specific problems are going to go away. There's all these

2489
02:44:03.120 --> 02:44:06.320
other problems that aren't going to go away, but we can point to the problems that are going to,

2490
02:44:06.320 --> 02:44:12.880
that are going to go away. So, this is, kind of, like, this is a defense against scope creep,

2491
02:44:13.760 --> 02:44:18.160
because when somebody says, yeah, but we also want to do this and this other thing, you say, yeah, but

2492
02:44:18.160 --> 02:44:21.840
we are making this, these problems are actually going to go away at the end of this four weeks.

2493
02:44:21.840 --> 02:44:24.400
Isn't that a win? And they're going to be, like, oh, yeah, that's actually a win.

2494
02:44:27.520 --> 02:44:29.040
That's good. That's helpful.

2495
02:44:33.120 --> 02:44:37.280
So, that's also, that's the kind of thing, too, where, like, sometimes it helps to be able to,

2496
02:44:37.280 --> 02:44:41.920
what I like about this loose format is I can, sometimes we can use language we wouldn't use

2497
02:44:41.920 --> 02:44:46.400
with the client that, kind of, cuts to the point a little bit more, you know, also just to remind

2498
02:44:47.280 --> 02:44:53.920
us about, like, what this is and stuff like that. So, very often I'll have just one of these in the

2499
02:44:53.920 --> 02:44:59.680
corner of a shaping document, and then if there is a bigger thing, usually we, kind of, don't need it

2500
02:44:59.680 --> 02:44:59.840
if we're

2501
02:45:00.000 --> 02:45:02.560
in a project already, this is kind of like,

2502
02:45:02.560 --> 02:45:05.800
but if we sometimes have to tease them apart,

2503
02:45:05.800 --> 02:45:07.760
this is just a quick, hacky way of doing that.

2504
02:45:07.760 --> 02:45:08.600
You know what I mean?

2505
02:45:08.600 --> 02:45:10.480
Of just kind of separating out,

2506
02:45:10.480 --> 02:45:12.040
what are the things that are high level

2507
02:45:12.040 --> 02:45:13.680
versus what are the things that are outcomes

2508
02:45:13.680 --> 02:45:16.480
for this specific slice here?

2509
02:45:19.360 --> 02:45:20.200
Yes.

2510
02:45:20.200 --> 02:45:23.560
Just for our context, like this would,

2511
02:45:23.560 --> 02:45:24.960
for our team, like would happen more

2512
02:45:24.960 --> 02:45:27.520
like in this, the blue, the discovery project.

2513
02:45:27.520 --> 02:45:29.080
Like I can see us, the like,

2514
02:45:29.080 --> 02:45:32.240
the outcome of that being a bunch of these little

2515
02:45:32.240 --> 02:45:33.920
problem outcome, problem outcome,

2516
02:45:33.920 --> 02:45:36.000
like stacking them and kind of separating

2517
02:45:36.000 --> 02:45:37.400
all these problems out.

2518
02:45:37.400 --> 02:45:40.200
And then working with the clients kind of figure out

2519
02:45:40.200 --> 02:45:43.160
like which of these do we want to even pursue?

2520
02:45:43.160 --> 02:45:45.720
And then turning that into like kind of packaging it

2521
02:45:45.720 --> 02:45:47.800
and making that our framed document.

2522
02:45:47.800 --> 02:45:50.600
Because I think it was, that was,

2523
02:45:50.600 --> 02:45:51.760
we were better because we like,

2524
02:45:51.760 --> 02:45:53.280
oh, we know breadboarding now,

2525
02:45:53.280 --> 02:45:54.800
or we know like we should use breadboarding.

2526
02:45:54.800 --> 02:45:57.640
So we jumped into it headfirst and try to,

2527
02:45:57.640 --> 02:45:58.640
and it was definitely helpful.

2528
02:45:58.640 --> 02:46:00.880
Like at least we were getting a lot more clear,

2529
02:46:00.880 --> 02:46:02.880
but it very quickly, I mean, it was,

2530
02:46:02.880 --> 02:46:05.840
it was very, it was such a knot of different things

2531
02:46:05.840 --> 02:46:06.680
that they wanted.

2532
02:46:06.680 --> 02:46:08.320
And then we were kind of guessing at what we think

2533
02:46:08.320 --> 02:46:09.160
the problem was.

2534
02:46:09.160 --> 02:46:10.880
So we kind of, we went a little too deep,

2535
02:46:10.880 --> 02:46:14.280
I think at that first dive.

2536
02:46:14.280 --> 02:46:15.640
Cause if I think we got it,

2537
02:46:16.960 --> 02:46:17.960
I think the, well, the good,

2538
02:46:17.960 --> 02:46:19.880
the benefits of it is that we went really deep

2539
02:46:19.880 --> 02:46:22.640
into like learning their process and learning the business.

2540
02:46:22.640 --> 02:46:23.600
And there's just some of that stuff

2541
02:46:23.600 --> 02:46:25.600
you just have to get the lay of the land

2542
02:46:25.600 --> 02:46:27.120
before you can even start to make,

2543
02:46:27.240 --> 02:46:30.400
you know, intelligent recommendations.

2544
02:46:30.400 --> 02:46:32.920
So, but I think this would have been a helpful tool.

2545
02:46:32.920 --> 02:46:35.880
I can see us using that a lot, you know,

2546
02:46:35.880 --> 02:46:38.040
just even just separating out these different things for us.

2547
02:46:38.040 --> 02:46:40.880
Cause we, we had them spread across different Figma boards

2548
02:46:40.880 --> 02:46:41.720
and different things.

2549
02:46:41.720 --> 02:46:44.120
It was just, it got very chaotic.

2550
02:46:44.120 --> 02:46:44.960
I remember that.

2551
02:46:44.960 --> 02:46:47.560
And then we try to distill it down to three projects.

2552
02:46:47.560 --> 02:46:50.000
So we didn't totally overwhelm them.

2553
02:46:50.000 --> 02:46:53.080
So that's my takeaway from that right there.

2554
02:46:58.080 --> 02:46:59.840
Bruce, did you get an answer to the question

2555
02:46:59.840 --> 02:47:01.200
about changing the frame?

2556
02:47:03.680 --> 02:47:07.920
No, but I think I'm, we're understanding, you know,

2557
02:47:07.920 --> 02:47:10.320
just to bring it back to that.

2558
02:47:10.320 --> 02:47:14.560
I think if, I think these, if we need to change it,

2559
02:47:14.560 --> 02:47:16.760
if we're super clear and then our team's super clear.

2560
02:47:16.760 --> 02:47:18.000
And when we talk to the client,

2561
02:47:18.000 --> 02:47:20.760
we're continuing to bring this up every single time

2562
02:47:20.760 --> 02:47:23.880
to remind them that's when they start to think about it.

2563
02:47:23.880 --> 02:47:24.880
And they start to like, oh,

2564
02:47:24.880 --> 02:47:26.520
but what if I asked this question, this question,

2565
02:47:27.040 --> 02:47:29.440
that helps us bring it back to a central point

2566
02:47:30.360 --> 02:47:33.000
because they may come back to us and say, no, like,

2567
02:47:33.000 --> 02:47:35.840
you know, well, this thing is way more important.

2568
02:47:35.840 --> 02:47:37.040
And at that point, like,

2569
02:47:37.040 --> 02:47:38.480
we also have to kind of hold the line.

2570
02:47:38.480 --> 02:47:41.600
Cause like we already started the project.

2571
02:47:41.600 --> 02:47:44.440
So we have to, that's where we have been thrown.

2572
02:47:44.440 --> 02:47:48.240
It's easy for clients to be like, we go 30% down one path.

2573
02:47:48.240 --> 02:47:50.480
And then they're like, oh no, this other thing over here.

2574
02:47:50.480 --> 02:47:52.120
And you end up like, you know,

2575
02:47:52.120 --> 02:47:55.120
sort of having this weird path up the mountain,

2576
02:47:55.120 --> 02:47:55.960
so to speak.

2577
02:47:55.960 --> 02:47:58.240
And it's just changing strategies like that

2578
02:47:58.240 --> 02:48:02.320
is a great way to just burn money and not see any outcome.

2579
02:48:02.320 --> 02:48:03.680
Oh, for sure, yeah.

2580
02:48:03.680 --> 02:48:04.520
So we want to kind of,

2581
02:48:04.520 --> 02:48:07.040
we're trying to like teach them something,

2582
02:48:07.040 --> 02:48:08.200
show this, the framework,

2583
02:48:08.200 --> 02:48:09.560
this is like the right way to do it,

2584
02:48:09.560 --> 02:48:12.080
but they have to kind of get on board with the process,

2585
02:48:12.080 --> 02:48:13.520
which they're, you know,

2586
02:48:13.520 --> 02:48:15.520
part of that discovery project is figuring out

2587
02:48:15.520 --> 02:48:16.360
if they're going to be okay

2588
02:48:16.360 --> 02:48:19.040
with this slightly different way of doing it.

2589
02:48:20.200 --> 02:48:21.960
And, but it is a more effective way

2590
02:48:21.960 --> 02:48:25.560
because we'll get to very meaningful, you know,

2591
02:48:25.560 --> 02:48:29.000
upgrades to their business and their platform as we go.

2592
02:48:30.200 --> 02:48:33.200
So yeah, anyway, that's, so yes,

2593
02:48:33.200 --> 02:48:35.720
I think we could change it if there's,

2594
02:48:35.720 --> 02:48:37.480
if there's like,

2595
02:48:37.480 --> 02:48:40.960
that means that we really understand their current phase.

2596
02:48:40.960 --> 02:48:42.120
We see the framed and they say,

2597
02:48:42.120 --> 02:48:43.720
actually we want to change this one bullet

2598
02:48:43.720 --> 02:48:44.920
to this other thing.

2599
02:48:44.920 --> 02:48:46.560
And now we have a very clear idea

2600
02:48:46.560 --> 02:48:47.760
of like what the current framework is.

2601
02:48:47.760 --> 02:48:49.680
And we're just really connecting this one thing

2602
02:48:49.680 --> 02:48:50.520
over here to that.

2603
02:48:50.520 --> 02:48:53.360
And it feels very, oh yes, that's like half a day.

2604
02:48:53.360 --> 02:48:56.280
We can make that one change and get them a ton more value.

2605
02:48:56.280 --> 02:49:00.240
I think that's where we have that as sort of the frame,

2606
02:49:00.240 --> 02:49:02.000
of course, for the project.

2607
02:49:02.000 --> 02:49:03.240
But once we get into the shaping,

2608
02:49:03.240 --> 02:49:05.840
we really have a grasp on every little dial.

2609
02:49:05.840 --> 02:49:07.960
And so that when they do want to tweak it,

2610
02:49:07.960 --> 02:49:08.960
we just have to be very clear.

2611
02:49:08.960 --> 02:49:12.920
What we can't do is be fuzzy on where all the things connect

2612
02:49:12.920 --> 02:49:15.000
and all the interdependencies.

2613
02:49:15.000 --> 02:49:17.280
And then they make it, you know, a question.

2614
02:49:17.280 --> 02:49:19.240
Then we jumped to like, yeah, sure we can do that.

2615
02:49:19.280 --> 02:49:22.000
And then now we've like expanded that frame,

2616
02:49:22.000 --> 02:49:23.760
but now it's not updated in the document.

2617
02:49:23.760 --> 02:49:25.960
So that's where expectations,

2618
02:49:25.960 --> 02:49:28.480
that's where I'm sensitive to.

2619
02:49:28.480 --> 02:49:30.360
Cause I'd rather like, hey, we did these four bullets

2620
02:49:30.360 --> 02:49:33.200
and we gave you the four bullets on time, on budget.

2621
02:49:33.200 --> 02:49:34.360
What do you want to do next?

2622
02:49:34.360 --> 02:49:36.240
You know, maybe those weren't the four bullets,

2623
02:49:36.240 --> 02:49:39.120
but the point is like the process executed

2624
02:49:39.120 --> 02:49:41.840
on those four bullets the right way.

2625
02:49:41.840 --> 02:49:42.680
Yeah.

2626
02:49:44.520 --> 02:49:49.040
Yeah, I've noticed that the things that we want

2627
02:49:49.840 --> 02:49:51.400
are very hard to keep fixed.

2628
02:49:51.400 --> 02:49:54.920
Like on the like sort of feature functionality side,

2629
02:49:54.920 --> 02:49:57.200
it's really hard to keep that under control.

2630
02:49:57.200 --> 02:50:00.280
But the things, the problems that we need to go away.

2631
02:50:00.000 --> 02:50:05.000
way, it's much easier to align on like, oh, this thing's first, or like, if that

2632
02:50:05.000 --> 02:50:06.920
problem goes away, that's a big success for us.

2633
02:50:07.240 --> 02:50:14.220
So, um, what I find is that like, there's the, there's the monster sandwich level

2634
02:50:14.220 --> 02:50:20.760
problem, which gets us into the relationship, but it's the, it's the piece

2635
02:50:20.760 --> 02:50:25.280
of the problem of like this particular piece, you know, is going away.

2636
02:50:25.520 --> 02:50:28.400
That's kind of the thing that really helps us to stay on the same track.

2637
02:50:28.400 --> 02:50:30.000
And we can keep referring back to that.

2638
02:50:31.880 --> 02:50:33.840
And it's very often missing, you know what I mean?

2639
02:50:33.840 --> 02:50:38.040
This, uh, this kind of what's wrong with the current way at this level, you

2640
02:50:38.040 --> 02:50:38.400
know what I mean?

2641
02:50:38.400 --> 02:50:42.480
So you can really see like the, the before after that's good.

2642
02:50:42.520 --> 02:50:42.880
Yeah.

2643
02:50:45.680 --> 02:50:48.360
Um, yeah, cool.

2644
02:50:48.680 --> 02:50:53.600
Um, uh, how are you guys feeling?

2645
02:50:54.200 --> 02:51:02.200
Um, so we, we, uh, we spent the first half, um, kind of detailing out the

2646
02:51:02.200 --> 02:51:07.200
current system and, uh, and then kind of, uh, sequencing the path through it in

2647
02:51:07.200 --> 02:51:09.400
terms of what problem to solve for a second, third.

2648
02:51:10.040 --> 02:51:12.680
Um, and then the second half, we kind of ended up a little bit

2649
02:51:12.680 --> 02:51:13.880
more on the framing side.

2650
02:51:14.800 --> 02:51:23.440
Um, yeah, any, um, uh, any takeaways from the second half

2651
02:51:23.960 --> 02:51:27.200
from the rest of you guys or anything outstanding also that's that, that we

2652
02:51:27.200 --> 02:51:29.480
didn't get to that, that's, that's a question hanging.

2653
02:51:42.680 --> 02:51:43.960
It's been pretty rich for me.

2654
02:51:44.920 --> 02:51:45.280
It's good.

2655
02:51:46.040 --> 02:51:46.200
Okay.

2656
02:51:47.720 --> 02:51:53.040
Uh, I mean, really enjoyed the framing part.

2657
02:51:53.040 --> 02:51:55.200
It almost makes me want to go into sales.

2658
02:51:58.040 --> 02:52:00.000
Almost not, not really.

2659
02:52:02.960 --> 02:52:07.200
The, uh, what, you know, I learned about, um, I was always

2660
02:52:07.200 --> 02:52:08.760
horrified by the idea of sales.

2661
02:52:09.080 --> 02:52:14.320
And, um, uh, I, I learned about it from, from Bob and, uh, in connection

2662
02:52:14.360 --> 02:52:16.920
with kind of a lot of this like sort of framing stuff, we didn't have the

2663
02:52:16.920 --> 02:52:20.720
word for it at the time, but, um, a lot of these like skills I learned from

2664
02:52:20.720 --> 02:52:27.000
Bob and, um, the thing that made it work for me is that like, he, he explains

2665
02:52:27.000 --> 02:52:31.680
sales, all of this stuff, like from the viewpoint of an engineer, which is why

2666
02:52:31.680 --> 02:52:35.160
I like it, like, cause it's like, you can, you can, you can, you can sort of

2667
02:52:35.160 --> 02:52:38.880
systematize like how it's working and why it works and when it breaks down.

2668
02:52:38.880 --> 02:52:39.600
And you know what I mean?

2669
02:52:39.600 --> 02:52:42.760
Like, and for me, that helps a lot to, to be able to rationalize it.

2670
02:52:42.760 --> 02:52:46.480
And it's not just the, uh, it's not just, um, you know, steaks, steak

2671
02:52:46.480 --> 02:52:49.200
dinners and stuff like that, but there's kind of like a lot of, uh,

2672
02:52:49.960 --> 02:52:53.120
there's a lot of, uh, function, something that you can analyze that,

2673
02:52:53.120 --> 02:52:55.040
that actually works in it, you know?

2674
02:52:56.120 --> 02:52:58.120
And, uh, Bob, are you talking about uncle Bob?

2675
02:52:58.920 --> 02:53:02.240
No, this is, no, this is, uh, this is Bob Mesta.

2676
02:53:02.800 --> 02:53:06.360
He's a, he's a really good friend and, uh, and mentor of mine.

2677
02:53:06.440 --> 02:53:13.080
Um, he, uh, uh, well, yeah, you, you, if you, uh, if you, if you look

2678
02:53:13.080 --> 02:53:14.600
them up, you can find all kinds of stuff.

2679
02:53:14.720 --> 02:53:16.440
Um, he's just right.

2680
02:53:17.720 --> 02:53:21.520
He's he's the, he's the original guy, um, behind, uh, jobs to

2681
02:53:21.520 --> 02:53:22.880
be done with Clay Christensen.

2682
02:53:23.880 --> 02:53:24.380
Yes.

2683
02:53:24.480 --> 02:53:25.480
I do remember that.

2684
02:53:25.480 --> 02:53:29.600
I think I've read some of the stuff, but he's actually a, he's a

2685
02:53:29.600 --> 02:53:31.640
hardcore engineer in his soul.

2686
02:53:31.640 --> 02:53:35.120
That's his real background and, uh, working in everything from

2687
02:53:35.160 --> 02:53:40.720
automotive to like NASA to, um, even like food, food production,

2688
02:53:41.200 --> 02:53:43.320
you know, like, uh, across industries.

2689
02:53:43.360 --> 02:53:45.760
And, uh, he's, he's a fascinating guy.

2690
02:53:45.800 --> 02:53:47.560
So yeah, one of my main mentors.

2691
02:53:48.400 --> 02:53:48.900
Nice.

2692
02:53:49.120 --> 02:53:49.620
That's great.

2693
02:53:49.620 --> 02:53:50.120
We'll check out.

2694
02:53:50.480 --> 02:53:50.980
Yeah.

2695
02:53:51.920 --> 02:53:56.200
Um, uh, anything, anything unclear about like next steps?

2696
02:53:58.920 --> 02:53:59.420
It's good.

2697
02:53:59.420 --> 02:54:04.000
I think, I mean, um, I think our next call is in a month.

2698
02:54:04.120 --> 02:54:08.280
Uh, so we'll see, we should be wrapping this up or have wrapped it up, but I

2699
02:54:08.280 --> 02:54:12.760
mean, I feel pretty good with all the tools, uh, you've given us all the tools

2700
02:54:12.760 --> 02:54:17.280
we need, we just need to go, you know, apply those again, kind of refresh it

2701
02:54:17.400 --> 02:54:22.800
and then have something, um, to just kind of, it's, you know, it's just resetting.

2702
02:54:22.800 --> 02:54:23.680
It's kind of taking this.

2703
02:54:23.680 --> 02:54:27.680
And once it's clear, like it's, it's very, you know, we can show that to the

2704
02:54:27.680 --> 02:54:32.340
client, it makes it easy, but it's really hard to, to fake clarity.

2705
02:54:32.360 --> 02:54:36.500
Like if we're vaguely, we're just filler busting, like making stuff up and it's

2706
02:54:36.500 --> 02:54:39.620
easy for any of us to do that when we don't really know what it is that we're

2707
02:54:39.620 --> 02:54:43.980
like, um, so we can be really clear and make it super concrete, then it's

2708
02:54:43.980 --> 02:54:45.060
just going to protect us.

2709
02:54:45.140 --> 02:54:48.460
Um, and also like, we're, this will be our second call with them.

2710
02:54:48.460 --> 02:54:49.860
I think the second check-in call.

2711
02:54:49.860 --> 02:54:54.320
So we'll just make sure we kind of course correct very quickly and like, Hey, just

2712
02:54:54.320 --> 02:54:54.820
bring it back.

2713
02:54:54.860 --> 02:54:56.060
Remember, this is the document.

2714
02:54:56.300 --> 02:54:57.540
Just want to reset expectation.

2715
02:54:57.540 --> 02:54:58.800
Now you guys are sending these things over here.

2716
02:54:58.800 --> 02:54:59.900
We got it, captured it.

2717
02:54:59.900 --> 02:55:00.400
So.

2718
02:55:00.000 --> 02:55:01.880
We're here on this other board, right.

2719
02:55:01.880 --> 02:55:03.480
Here's like where we're going to focus.

2720
02:55:03.480 --> 02:55:06.820
This call is like where these specific three things are or wherever.

2721
02:55:07.180 --> 02:55:13.640
Um, and, uh, you know, I think also once it's very clear for us, we can also kind

2722
02:55:13.640 --> 02:55:15.940
of see a quicker path to like test this.

2723
02:55:16.000 --> 02:55:21.820
And so, you know, even just grabbing, you know, 20, I think I, we talked about this

2724
02:55:21.900 --> 02:55:25.080
early, I don't know if this got done or not, but taking 20 files, throwing it

2725
02:55:25.080 --> 02:55:29.400
into a quick AI thing and just giving it as the full context window, like all

2726
02:55:29.400 --> 02:55:33.240
these files and just see how good it is just right off the bat without

2727
02:55:33.240 --> 02:55:34.980
us doing anything fancy to it.

2728
02:55:35.360 --> 02:55:40.240
Um, you know, that's something we could do as a proof of concept quickly.

2729
02:55:40.660 --> 02:55:45.120
Um, and those are things where I see like rapid prototyping kind of just to quickly,

2730
02:55:45.140 --> 02:55:49.080
cause we, a lot of things to see, and that would also help clarify like the vision.

2731
02:55:49.520 --> 02:55:53.840
Um, that's, that's even something we could probably hack before the call.

2732
02:55:53.840 --> 02:55:57.540
But, um, yeah, just cause I think, I think when they start to, I mean, a

2733
02:55:57.540 --> 02:56:02.180
demo is worth a thousand words, like showing a two minute demo, something,

2734
02:56:02.540 --> 02:56:04.020
um, clarifies all that stuff.

2735
02:56:04.540 --> 02:56:06.220
So, um, yep.

2736
02:56:06.660 --> 02:56:07.500
Were you gonna say something Ira?

2737
02:56:08.900 --> 02:56:10.540
Oh, no, I was just, um, okay.

2738
02:56:10.540 --> 02:56:12.060
I thought you unmuted, but that's good.

2739
02:56:12.260 --> 02:56:12.460
Cool.

2740
02:56:13.200 --> 02:56:15.980
Uh, I love, I love, I love what you said about like this, like you can't fake

2741
02:56:15.980 --> 02:56:19.540
clarity, like the quote I usually think of is that it's, it's, it's, uh, it's

2742
02:56:19.540 --> 02:56:26.940
easy to macro bullshit and hard to micro bullshit, you know, like, so it's like.

2743
02:56:27.780 --> 02:56:28.420
Yeah.

2744
02:56:29.020 --> 02:56:30.060
Um, cool.

2745
02:56:30.100 --> 02:56:30.580
Okay.

2746
02:56:30.620 --> 02:56:31.140
All right, guys.

2747
02:56:31.180 --> 02:56:32.460
Hey, excellent.

2748
02:56:32.700 --> 02:56:33.660
I mean, nice session.

2749
02:56:33.980 --> 02:56:34.500
Nice work.

2750
02:56:35.500 --> 02:56:38.460
And, uh, uh, you know, good luck with the project and we'll be in touch.

2751
02:56:39.220 --> 02:56:39.660
Thanks, Ryan.

2752
02:56:39.660 --> 02:56:40.100
Appreciate it.

2753
02:56:41.220 --> 02:56:41.540
All right.

2754
02:56:42.260 --> 02:56:42.700
See you guys.

2755
02:56:42.700 --> 02:56:43.180
Have a good one.

2756
02:56:43.220 --> 02:56:43.700
Cheers.