Meeting Title: AI Service Standup Date: 2026-01-20 Meeting participants: Samuel Roberts, Uttam Kumaran, Mustafa Raja, Casie Aviles, Pranav Narahari, Amber Lin, Gabriel Lam, Clarence Stone


WEBVTT

1 00:01:24.710 00:01:25.810 Uttam Kumaran: Hello.

2 00:01:27.820 00:01:28.560 Samuel Roberts: Ayy.

3 00:01:29.270 00:01:30.210 Uttam Kumaran: Hey, dude.

4 00:01:31.370 00:01:32.670 Samuel Roberts: We were busy last night.

5 00:01:33.210 00:01:39.660 Uttam Kumaran: I know, I’m so… dude, yesterday was, like, maybe, like, the most normal workday I’ve had in, like, 2 years.

6 00:01:39.660 00:01:40.169 Samuel Roberts: That’s great.

7 00:01:40.170 00:01:41.699 Uttam Kumaran: I felt like I was, like…

8 00:01:42.320 00:01:45.040 Uttam Kumaran: I felt like I was just, like, on the team, like…

9 00:01:46.150 00:01:54.919 Uttam Kumaran: Because I was like, oh, I have, like, time to, like, eat. I, like, went to get a haircut, but, like, I still did a bunch of work in the morning, like…

10 00:01:55.570 00:01:57.990 Uttam Kumaran: And then, yeah, I just,

11 00:01:58.850 00:02:04.020 Uttam Kumaran: I’ve been waiting to sort of be able to have, like, a couple free hours to just push on…

12 00:02:04.390 00:02:14.910 Uttam Kumaran: getting my mind around how to use codecs, and then how to use cursor codecs, and, like, I wanted to learn more about deployment,

13 00:02:15.760 00:02:21.749 Uttam Kumaran: Yeah, so it was… it’s awesome. This is great. Like, this is, like, exactly what I hoped it was gonna be like. Like, it’s not…

14 00:02:21.750 00:02:22.150 Samuel Roberts: Sweet.

15 00:02:22.150 00:02:24.660 Uttam Kumaran: It’s not straightforward, and I can send,

16 00:02:25.170 00:02:31.269 Uttam Kumaran: What I’ll probably also start to do is save my cursor interaction somewhere.

17 00:02:31.270 00:02:32.030 Samuel Roberts: Oh, yeah.

18 00:02:32.030 00:02:34.520 Uttam Kumaran: Of, like, how I accomplished that.

19 00:02:34.730 00:02:37.739 Uttam Kumaran: That way, eventually.

20 00:02:38.380 00:02:43.300 Uttam Kumaran: More and more people will be using Kirscher to ship stuff, and we’ll have some insights into that.

21 00:02:43.930 00:02:48.379 Uttam Kumaran: You know, into, like, what they’re doing, and how do we make things more streamlined, but yeah, it was fine.

22 00:02:48.740 00:02:49.519 Samuel Roberts: Right, right.

23 00:02:50.610 00:02:51.360 Samuel Roberts: Cool.

24 00:02:51.790 00:02:54.339 Samuel Roberts: Yeah, like, Codex in Cursor, how is that?

25 00:02:55.400 00:02:57.620 Uttam Kumaran: It was fine, like, yeah, I,

26 00:02:58.060 00:03:05.880 Uttam Kumaran: I just wanted to see the differences in the interactions between that and Opus. I mean, it still acts as an agent, but it sort of, like, loops…

27 00:03:06.060 00:03:08.800 Uttam Kumaran: And does a bunch of things, versus, like…

28 00:03:09.150 00:03:11.730 Uttam Kumaran: One thing at a time.

29 00:03:13.790 00:03:20.039 Uttam Kumaran: It was good. I mean, I’m feeling more and more confident on the fact that, like, if I just had, like, my…

30 00:03:21.180 00:03:25.630 Uttam Kumaran: Based prompt, my, like, requirements actually set a little bit better.

31 00:03:25.890 00:03:35.400 Uttam Kumaran: I could have, probably handed the whole thing to Codex, like, the UI. And actually, in general now, I also enabled,

32 00:03:35.800 00:03:38.600 Uttam Kumaran: I enabled codecs within linear.

33 00:03:38.840 00:03:39.560 Uttam Kumaran: So…

34 00:03:39.560 00:03:40.479 Samuel Roberts: Yeah, I saw that.

35 00:03:41.080 00:03:42.940 Uttam Kumaran: I’m gonna see…

36 00:03:43.380 00:03:54.819 Uttam Kumaran: today, for some of the… I’m basically gonna try to do, like, sort of, like, the ultimate flow, which is, like, I’m gonna work with Cursor to write, like, the PRD.

37 00:03:54.990 00:04:02.390 Uttam Kumaran: And then I’m going to have it break that up into tickets, and then I’m gonna…

38 00:04:03.350 00:04:06.409 Uttam Kumaran: Kind of just, like, a sign out, like.

39 00:04:06.820 00:04:08.579 Uttam Kumaran: I guess, one by one, to curse.

40 00:04:08.580 00:04:09.050 Samuel Roberts: Yeah.

41 00:04:09.050 00:04:10.440 Uttam Kumaran: Took to Codex.

42 00:04:10.730 00:04:14.649 Uttam Kumaran: And see if it… like, see where it kind of breaks.

43 00:04:14.790 00:04:17.040 Uttam Kumaran: Yeah.

44 00:04:18.700 00:04:19.350 Samuel Roberts: Sweet.

45 00:04:22.290 00:04:28.949 Samuel Roberts: So you… I just… I mean, we can get… I guess we can start the meeting now that everyone’s here, but do you tend to use Opus specifically in Cursor? Is that how you…

46 00:04:29.510 00:04:30.010 Samuel Roberts: Have you been.

47 00:04:30.010 00:04:40.159 Uttam Kumaran: Yeah, I typically will use Opus. I need to check, like, if we’re getting, like, absolutely slammed on the billing side by that. Yeah. I’m probably, like, overusing it.

48 00:04:40.490 00:04:47.799 Uttam Kumaran: But, yeah, I typically use Opus, and then I switch to Sonnet for writing, or I switch to Auto for, like, execution.

49 00:04:47.800 00:04:48.600 Samuel Roberts: Yeah, okay.

50 00:04:49.470 00:04:58.590 Samuel Roberts: Yeah, I’ve been using a lot of auto for execution. I really… I was playing around with the composer for a while, which was so fast, and the code was pretty good a lot of the time.

51 00:04:59.620 00:05:00.380 Samuel Roberts: Yeah, that’s a…

52 00:05:00.380 00:05:05.999 Uttam Kumaran: I think for me, like, the code quality side, more of what I’m gonna ask it to do is, like, do unit tests.

53 00:05:06.000 00:05:06.770 Samuel Roberts: Yeah.

54 00:05:06.800 00:05:08.600 Uttam Kumaran: And like…

55 00:05:10.090 00:05:14.569 Uttam Kumaran: Yeah, I’m gonna try to work on some quality of life stuff for that repo in general.

56 00:05:14.890 00:05:15.920 Samuel Roberts: Sorry.

57 00:05:17.740 00:05:19.720 Samuel Roberts: Alright, well, welcome, everyone.

58 00:05:21.590 00:05:22.949 Samuel Roberts: How’s everyone doing?

59 00:05:26.420 00:05:28.050 Gabriel Lam: Doing alright?

60 00:05:28.050 00:05:28.720 Samuel Roberts: Cool.

61 00:05:30.640 00:05:36.409 Samuel Roberts: Cool, cool. Yeah, I’m moving a little slow this morning, and my kid was up a bunch last night, so forgive me if I’m a little,

62 00:05:37.010 00:05:38.420 Samuel Roberts: Still waking up a little bit.

63 00:05:39.060 00:05:43.780 Samuel Roberts: Yeah, I guess let’s jump in, let’s talk ABC.

64 00:05:49.100 00:05:49.850 Casie Aviles: Yeah.

65 00:05:50.170 00:05:53.099 Samuel Roberts: Yeah, if you want to jump on, or jump to it?

66 00:05:53.100 00:05:58.499 Casie Aviles: I can start, for ABC, there’s, I didn’t really get back.

67 00:05:58.900 00:06:01.320 Casie Aviles: Or, like, I mean, I didn’t get anything back from…

68 00:06:01.910 00:06:02.310 Samuel Roberts: Okay.

69 00:06:02.310 00:06:07.420 Casie Aviles: genes yet, I’m thinking. I’m not sure if, like, if it’s the best…

70 00:06:08.000 00:06:12.869 Casie Aviles: If it’s best to just get it out, and then just make sure to be,

71 00:06:13.590 00:06:17.299 Casie Aviles: To kind of, you know, it’s, like, going to be an assisted launch, where…

72 00:06:17.760 00:06:23.439 Casie Aviles: It’s going to be live, but we’ll continue to monitor and maintain things.

73 00:06:23.960 00:06:25.250 Samuel Roberts: Okay.

74 00:06:25.550 00:06:31.830 Casie Aviles: Because I really do think that with the admin UI out, it should help us with the…

75 00:06:32.780 00:06:38.940 Casie Aviles: the data freshness issues that, that the CSRs are…

76 00:06:39.250 00:06:45.370 Casie Aviles: Sending and flagging to us, so that’s… that’s, like, one path that I’m considering.

77 00:06:46.390 00:06:50.480 Casie Aviles: Otherwise, we could… I could just keep, like.

78 00:06:51.030 00:06:55.329 Casie Aviles: triaging the tickets and doing it manually on Superbase.

79 00:06:56.100 00:06:57.180 Samuel Roberts: Yeah, that seems…

80 00:06:57.570 00:07:02.220 Samuel Roberts: But… It’s unsustainable, but yeah, I mean, if there’s… if she hasn’t gotten back to you yet, I would say…

81 00:07:02.710 00:07:08.020 Samuel Roberts: It’s not a bad idea to flip that on and start using it, at least, and we can still make changes to it if she has…

82 00:07:08.440 00:07:12.019 Samuel Roberts: If she has thoughts, I was just hoping to get that sooner rather than later, but…

83 00:07:12.450 00:07:15.239 Samuel Roberts: If she’s not gotten back, that’s fine.

84 00:07:17.000 00:07:17.660 Casie Aviles: Okay.

85 00:07:17.830 00:07:24.480 Casie Aviles: Yeah, I think I can move towards that, and it’s also just going to make it easier for us and more efficient to

86 00:07:24.730 00:07:28.890 Casie Aviles: You know, address those… Tickets in bulk.

87 00:07:28.890 00:07:30.330 Samuel Roberts: Yes, definitely.

88 00:07:32.320 00:07:38.969 Samuel Roberts: Cool, okay, yeah, so yeah, I was gonna go ahead with that. That’s gonna be fine.

89 00:07:38.970 00:07:45.519 Casie Aviles: That… even for the tech zip codes here on… on… that are on… This linear view.

90 00:07:45.520 00:07:46.100 Samuel Roberts: Yep.

91 00:07:46.100 00:07:51.049 Casie Aviles: That should be addressable once we have that live and up.

92 00:07:53.510 00:07:56.619 Samuel Roberts: Okay, yeah, I think that sounds good. We’ll say focus on that today, then.

93 00:07:57.950 00:07:58.590 Casie Aviles: Okay.

94 00:07:58.800 00:08:03.499 Casie Aviles: Yeah, I’ll make sure that we have the updated one, and I’ll just let them know.

95 00:08:04.040 00:08:05.470 Samuel Roberts: Okay. Cool.

96 00:08:07.430 00:08:13.160 Mustafa Raja: Yeah, from my side, I worked on setting up, the evals locally.

97 00:08:13.760 00:08:22.030 Mustafa Raja: I’m currently working on that right now, so, I’ll be testing that locally after this meeting, and, probably come up with a loom.

98 00:08:22.320 00:08:23.350 Samuel Roberts: Premier.

99 00:08:25.650 00:08:32.290 Mustafa Raja: For the magnet cancellations, I believe, I would need a review from Amber.

100 00:08:32.289 00:08:34.380 Samuel Roberts: To see if we can close this now.

101 00:08:35.130 00:08:41.320 Mustafa Raja: And the rest, I believe, established evaluation data set is now good to be done.

102 00:08:42.419 00:08:43.759 Samuel Roberts: Yes, that’s true.

103 00:08:48.290 00:08:52.519 Mustafa Raja: And then for, for the, evals that I am working on.

104 00:08:52.720 00:08:56.689 Mustafa Raja: Setting up evals, we would need a new ticket.

105 00:08:57.210 00:08:58.580 Mustafa Raja: Based on the Gantt chart.

106 00:09:01.830 00:09:04.140 Samuel Roberts: Yes. Okay.

107 00:09:04.640 00:09:06.590 Samuel Roberts: We have that in the Gantt, specifically?

108 00:09:06.590 00:09:08.340 Mustafa Raja: Yeah, I think a.

109 00:09:08.530 00:09:10.219 Samuel Roberts: Okay, I’ll put that in.

110 00:09:13.310 00:09:28.820 Pranav Narahari: Mustafa, on that, is there anything… because I know… so I, I wrote up a Notion doc on just, like, the core eval metrics. Is there anything that we need to sync on with,

111 00:09:29.020 00:09:30.390 Pranav Narahari: With what you’re working on?

112 00:09:31.720 00:09:44.900 Mustafa Raja: Yeah, so, eventually, so what you’re working on would be scorers, right? How should we score each evaluation, right? So currently, what I am doing is I’m just going to see how relevant it is with the output.

113 00:09:45.100 00:09:56.079 Mustafa Raja: And, I’m going to see what you are suggesting for scoring these, so we can ultimately have your thing implemented, rather than a built-in

114 00:09:56.280 00:09:57.660 Mustafa Raja: scooter.

115 00:09:58.020 00:09:59.179 Pranav Narahari: Okay, gotcha.

116 00:09:59.180 00:10:04.719 Mustafa Raja: Yeah, so send over the Notion doc to me, if I have any questions, I’ll…

117 00:10:04.890 00:10:07.080 Mustafa Raja: I’ll ping you, and we can meet on that.

118 00:10:07.080 00:10:11.010 Pranav Narahari: Sure, yeah, it’s in that, ticket. Okay, okay.

119 00:10:11.010 00:10:11.359 Mustafa Raja: It’s like…

120 00:10:11.360 00:10:11.910 Pranav Narahari: Let’s…

121 00:10:12.330 00:10:13.270 Samuel Roberts: Okay, cool.

122 00:10:16.230 00:10:20.540 Samuel Roberts: Great. Any other, updates there?

123 00:10:21.520 00:10:23.240 Mustafa Raja: Well, that’s what you mentioned from my side.

124 00:10:23.540 00:10:24.220 Samuel Roberts: Okay.

125 00:10:25.280 00:10:31.920 Amber Lin: I checked the Gantt chart, I added the tickets that we’re supposed to be doing this week,

126 00:10:32.060 00:10:41.930 Amber Lin: Based on this view, is this a… are we able to accomplish all of these? I feel like we haven’t, closed out the tickets from the last week.

127 00:10:41.930 00:10:45.219 Samuel Roberts: Yeah, there’s a few still outstanding from last week that,

128 00:10:45.840 00:10:50.789 Samuel Roberts: You should be able to close out soon, so we might be a little behind here,

129 00:10:52.930 00:10:56.899 Samuel Roberts: You can use eval logging…

130 00:10:59.930 00:11:03.860 Samuel Roberts: Yeah, we might need to sync on that and make sure that we adjust properly.

131 00:11:04.070 00:11:08.379 Pranav Narahari: I can help on some stuff, especially if,

132 00:11:08.610 00:11:14.399 Pranav Narahari: the Notion doc that I created doesn’t really need much, additional effort.

133 00:11:14.850 00:11:16.689 Pranav Narahari: Yeah, so, let me know, guys.

134 00:11:17.580 00:11:22.390 Samuel Roberts: Okay, yeah, that might be good. I would say…

135 00:11:22.390 00:11:25.300 Amber Lin: Anything on Casey’s plate that we can…

136 00:11:25.300 00:11:31.659 Samuel Roberts: Yeah, I mean, I think we can… these will hopefully… It’s sorted today, then.

137 00:11:32.610 00:11:36.210 Samuel Roberts: Right, like, the first 3 here, we just launched that out.

138 00:11:37.670 00:11:40.229 Samuel Roberts: I do it.

139 00:11:40.230 00:11:48.030 Casie Aviles: It’s not going to be a lot… much left to do here, it’s just really getting it live, you know, for those existing tickets.

140 00:11:48.030 00:11:48.900 Samuel Roberts: Right.

141 00:11:50.090 00:11:51.110 Samuel Roberts: Okay.

142 00:11:51.840 00:11:55.500 Samuel Roberts: Yeah, let me take a look at the Gantt, because, like, this might not be…

143 00:11:56.160 00:11:59.740 Samuel Roberts: We can probably kick this down, or move this to someone.

144 00:12:00.910 00:12:07.550 Samuel Roberts: It’s edge cases and regression. We still have to identify that, but that’ll happen with Mustafa’s work.

145 00:12:10.150 00:12:13.459 Samuel Roberts: Okay, yeah, let me take a look at the Gantt after this, and I’ll sync with you, Amber.

146 00:12:15.860 00:12:17.200 Amber Lin: Cool, sounds great.

147 00:12:17.200 00:12:18.720 Samuel Roberts: Okay, cool.

148 00:12:19.960 00:12:22.489 Samuel Roberts: Anything else before we move on to Leela?

149 00:12:26.420 00:12:27.150 Samuel Roberts: Cool.

150 00:12:27.660 00:12:28.890 Casie Aviles: I think that’s all for me.

151 00:12:29.130 00:12:39.440 Samuel Roberts: Great. Alright, so, updates on Lilo… I think we’re…

152 00:12:39.550 00:12:45.320 Samuel Roberts: We have been a banana in dev, we want them to play around with it a little bit, make sure it’s good, and then we can pretty much

153 00:12:45.480 00:12:48.929 Samuel Roberts: Promote everything all the way up and set up production today.

154 00:12:49.220 00:12:50.030 Uttam Kumaran: Cool.

155 00:12:50.030 00:12:53.069 Samuel Roberts: I think that’s… that’s the plan, and then…

156 00:12:53.200 00:13:00.920 Samuel Roberts: Once that’s all good and running, we can start to work with them on the things that they pushed.

157 00:13:01.630 00:13:04.950 Samuel Roberts: the front-end stuff that they pushed, I saw a few things in there.

158 00:13:05.140 00:13:10.099 Samuel Roberts: And then the other side of it is the Phase 2 stuff with,

159 00:13:11.800 00:13:15.570 Samuel Roberts: The warehouse setup and data on the forecasting.

160 00:13:15.740 00:13:20.419 Samuel Roberts: So, we may need to… Talk through that stuff a little bit.

161 00:13:22.520 00:13:34.510 Pranav Narahari: Yeah, so I created these tickets yesterday, the data warehouse setup, yeah, I specifically just assigned it to you, just so you can kind of help with, like, the… kind of the design of that. And then I can…

162 00:13:34.850 00:13:38.460 Pranav Narahari: Probably break that up into, like, a few more tickets to, like…

163 00:13:38.510 00:13:39.460 Samuel Roberts: Yeah.

164 00:13:39.530 00:13:42.599 Pranav Narahari: Yeah, to, like, you know, split up the load for that.

165 00:13:43.970 00:13:50.580 Pranav Narahari: Do you think we should, like, maybe have, like, a… like, an engineering sync on that, me, you, Casey?

166 00:13:51.180 00:13:56.749 Pranav Narahari: To just kind of, like, spit… spitball some ideas, or… you can just, like, talk us through, like, what you’re thinking about.

167 00:13:57.020 00:14:06.919 Samuel Roberts: Yeah, I thought… weren’t we gonna kick this down a little bit, and we were gonna do just the front end, basically, with the data coming in from the API for now? Or some, like, with, like…

168 00:14:07.150 00:14:13.330 Samuel Roberts: Input data, was that the plan originally? Or, not originally, but… We updated Phase 2.

169 00:14:14.420 00:14:20.919 Pranav Narahari: That’s not what the Gantt shows right now, but Utam, is that kind of… is the Gantt…

170 00:14:21.250 00:14:30.379 Uttam Kumaran: Yeah, phase two is they, like, there’s actually… we basically are recreating that… that, like, linear forecast spreadsheet that they sent.

171 00:14:31.080 00:14:34.049 Uttam Kumaran: I don’t know if you guys recall the one they shared.

172 00:14:34.050 00:14:39.810 Samuel Roberts: Yeah, yeah, I just wasn’t sure. I mean, we had talked about trying to speed that up, and whether or not the data warehouse was…

173 00:14:40.750 00:14:44.549 Samuel Roberts: critical for that. I thought it was just going to be, like, calculations, basically, and they were going to input data.

174 00:14:44.550 00:14:52.530 Uttam Kumaran: It is populations, I guess, like, what, what basically it is, is, like, we have to pull… we have to pull data and, like, do…

175 00:14:52.680 00:15:03.079 Uttam Kumaran: some transformation, and then represent it on the front end. I don’t… I guess, like, if I was to say, like, how that would work, basically, we need to…

176 00:15:03.760 00:15:08.129 Uttam Kumaran: Have a data warehouse, get, like, basically get, like, each brand’s…

177 00:15:08.250 00:15:11.020 Uttam Kumaran: Shopify data in there, and then…

178 00:15:11.160 00:15:18.800 Uttam Kumaran: all… or whatever the data is, all it’s doing is, like, running a SQL query on top of that to create, like, a summary table.

179 00:15:19.360 00:15:22.040 Uttam Kumaran: And then it’s representing that in the front end.

180 00:15:22.990 00:15:27.639 Samuel Roberts: you know, to do that via the API, like, I don’t know whether you can…

181 00:15:27.640 00:15:28.770 Uttam Kumaran: do…

182 00:15:29.240 00:15:37.920 Uttam Kumaran: all of that on the fly. I mean, the data warehouse and, like, moving the data in via AirByte is… they’re just doing, like, API calls to move the data.

183 00:15:37.920 00:15:38.830 Samuel Roberts: Right, right.

184 00:15:38.830 00:15:44.729 Uttam Kumaran: But you’re not… you don’t have to do it all the time, and it may be… it may be, like, millions of rows.

185 00:15:45.300 00:15:46.100 Samuel Roberts: Right.

186 00:15:48.250 00:15:54.160 Samuel Roberts: Okay. Okay, that makes sense. Yeah, I just… I couldn’t remember how we… I thought we were trying to speed it up, and I couldn’t remember exactly how we were.

187 00:15:54.330 00:16:03.290 Uttam Kumaran: Yeah, so the way we ride at speeding it up was actually just, like, simplifying what the output is, like, it’s not, like, some advanced forecasting cohort thing.

188 00:16:03.290 00:16:03.849 Samuel Roberts: Got it.

189 00:16:03.850 00:16:08.860 Uttam Kumaran: It’s actually just, like, a simple, linear… Forecast, basically.

190 00:16:08.860 00:16:09.490 Samuel Roberts: Okay.

191 00:16:09.880 00:16:16.639 Uttam Kumaran: So, I don’t know, Pranavi, if you still have access to… the, like…

192 00:16:16.810 00:16:22.550 Uttam Kumaran: Basically, the sheet that they… they shared. Right.

193 00:16:23.060 00:16:25.170 Uttam Kumaran: But that’s all we’re trying to replicate.

194 00:16:25.400 00:16:29.140 Samuel Roberts: Okay. Yeah, let’s, definitely sync that later. If we can find that sheet.

195 00:16:29.180 00:16:30.180 Pranav Narahari: Yeah.

196 00:16:30.450 00:16:34.240 Samuel Roberts: Share that out, and then figure out exactly what we need to pull.

197 00:16:34.360 00:16:39.219 Samuel Roberts: How much that’s gonna be, and see what we need to do. If we… if we can just, you know.

198 00:16:39.410 00:16:45.120 Samuel Roberts: do it through the API and do a basic quick sync, or if we actually need something like AirByte right now or not.

199 00:16:46.270 00:16:51.520 Pranav Narahari: And, the timeline on all of these items, right, is…

200 00:16:52.050 00:17:00.570 Pranav Narahari: 3 Fridays from now. So, this isn’t, like, the data warehouse setup isn’t something necessarily that we need to get done, like, this week.

201 00:17:00.610 00:17:06.410 Samuel Roberts: Right. Yeah, so just keep that in mind, too. Okay, yeah, I mean, we gotta plan that a little bit anyway, so that’ll be good. Okay.

202 00:17:07.400 00:17:13.970 Uttam Kumaran: So if you guys chat, and then if you… or if you want to let me know, I can hop on and explain, but, like, roughly.

203 00:17:14.109 00:17:22.149 Uttam Kumaran: like, the suggestion initially was to use a tool like AirByte, move in data… Shopify into a warehouse?

204 00:17:22.310 00:17:28.050 Uttam Kumaran: put some, like, light transformation on it, and then represent it on the front end. And so…

205 00:17:28.359 00:17:32.290 Uttam Kumaran: I believe that’s all gonna be fairly…

206 00:17:32.830 00:17:34.969 Uttam Kumaran: you know, possible. I don’t think it’s…

207 00:17:35.510 00:17:39.319 Uttam Kumaran: I don’t think it’s, like, too complex, given, like, what I saw in that spreadsheet.

208 00:17:39.320 00:17:48.310 Samuel Roberts: No, I think you’re right, I think you’re right. The question is, like, how, yeah, how complex do we need to make it? Is it… is, like, AirByte, or are we able to just… even for, like, the proof of concept, just get it?

209 00:17:48.580 00:17:52.170 Samuel Roberts: Out and make sure we know what the data is and everything.

210 00:17:53.830 00:17:56.270 Samuel Roberts: Over the next 3 weeks, that should be plenty of…

211 00:17:57.450 00:18:01.120 Samuel Roberts: I have time to sort out the best way to do that, and do it in phases, even.

212 00:18:01.720 00:18:03.999 Samuel Roberts: Phases within phases, so…

213 00:18:04.000 00:18:09.389 Uttam Kumaran: The thing is, like, they’re gonna quickly want to combine this data with other data.

214 00:18:09.390 00:18:10.549 Samuel Roberts: Yeah, you’re right.

215 00:18:10.840 00:18:16.319 Uttam Kumaran: And it’s, like, 80… like, I just don’t know whether you’re gonna be able to pull…

216 00:18:17.080 00:18:23.620 Uttam Kumaran: I just don’t know how, like, the current MCP works, but, like, for example, the way we typically handle it on the data side.

217 00:18:23.720 00:18:25.880 Uttam Kumaran: Is we,

218 00:18:26.420 00:18:39.419 Uttam Kumaran: we’re ingesting, like, all of the Shopify orders for some brands that may be, like, millions of orders, or it could be, like, hundreds of thousands of orders. We’re running, like, a SQL query that, like.

219 00:18:39.610 00:18:48.909 Uttam Kumaran: Sums the orders, like, sums the revenue from them, and then also can take an input and, like, forecast it forward.

220 00:18:49.020 00:18:53.420 Uttam Kumaran: you know, that’s like a typical SQL query. It’s just…

221 00:18:53.640 00:18:57.840 Uttam Kumaran: And then, that way, also, we’re only, doing that

222 00:18:58.050 00:19:02.279 Uttam Kumaran: Like, once, like, and we can rerun that model once per day.

223 00:19:02.440 00:19:04.259 Samuel Roberts: Sure. Okay.

224 00:19:04.470 00:19:08.310 Uttam Kumaran: Versus, like, having to pull the API, like, every time, like, I just… Sure.

225 00:19:08.310 00:19:13.509 Samuel Roberts: Yeah, no, I definitely… I definitely think that’s… that makes sense. So, the… the forecasting happens in SQL as well?

226 00:19:14.350 00:19:15.500 Uttam Kumaran: Yeah, typically.

227 00:19:15.500 00:19:16.110 Samuel Roberts: Okay.

228 00:19:16.600 00:19:17.759 Samuel Roberts: Cool, alright, yeah, let’s…

229 00:19:17.760 00:19:19.370 Uttam Kumaran: I mean, but again, I just, like, if…

230 00:19:19.480 00:19:25.160 Uttam Kumaran: I don’t know what the alternative is, like, if you think, hey, it’s actually possible for us to do this all

231 00:19:26.060 00:19:32.440 Uttam Kumaran: like, right now, a lot of companies, what they’re doing is they’re using, like, DuckDB, and they’re doing it, like, in the browser.

232 00:19:32.750 00:19:33.330 Samuel Roberts: Yeah.

233 00:19:33.560 00:19:37.690 Uttam Kumaran: So I just don’t know, I just don’t know what that world looks like, so…

234 00:19:37.690 00:19:38.530 Samuel Roberts: Okay.

235 00:19:38.530 00:19:41.709 Uttam Kumaran: to think of that alternative, because I know, sort of, like, what our typical

236 00:19:41.840 00:19:48.969 Uttam Kumaran: world is where we ETL all of that in, and then we run a model every day, and then we just display that.

237 00:19:48.970 00:19:50.110 Samuel Roberts: Yeah, that…

238 00:19:50.110 00:19:50.770 Uttam Kumaran: Yeah.

239 00:19:51.020 00:19:58.350 Samuel Roberts: That makes the most sense, definitely, like, long-term. Like, they’re gonna have a ton of brands, and the brands are all gonna have multiple sources and everything. Just trying to… yeah, okay.

240 00:19:59.010 00:20:03.379 Samuel Roberts: we gotta get to that, kind of, as… as cleanly as possible, I think.

241 00:20:04.510 00:20:05.340 Uttam Kumaran: Okay.

242 00:20:05.340 00:20:07.720 Samuel Roberts: Yeah, let’s sync on that later.

243 00:20:08.400 00:20:14.960 Samuel Roberts: Talk through some of that stuff, maybe do a quick, investigation over a few things before that, so maybe this afternoon would be good.

244 00:20:14.960 00:20:27.610 Pranav Narahari: Yeah, and I would say, so Bobby sent a message this morning. I think the number one priority is for me to get these Slack messages, like, the data discrepancy issue, like, sorted, and then also just nano Banana.

245 00:20:27.770 00:20:31.910 Samuel Roberts: Yeah, no, 100%. The goal today is definitely get everything into prod.

246 00:20:31.910 00:20:32.920 Pranav Narahari: Yeah.

247 00:20:33.500 00:20:38.309 Samuel Roberts: But yes, yes, Slack messages and then a banana are the… The main things we will…

248 00:20:38.950 00:20:42.950 Samuel Roberts: Get out today, but let’s definitely have some conversations about the other stuff.

249 00:20:43.170 00:20:43.720 Pranav Narahari: Cool.

250 00:20:43.940 00:20:44.929 Pranav Narahari: Sounds good.

251 00:20:44.930 00:20:45.890 Samuel Roberts: Alright, cool.

252 00:20:46.150 00:20:50.539 Samuel Roberts: Yeah, I would say grab some time this afternoon, if there’s something open.

253 00:20:50.890 00:20:51.610 Pranav Narahari: Okay.

254 00:20:51.610 00:20:52.260 Samuel Roberts: Thanks.

255 00:20:52.670 00:20:56.899 Samuel Roberts: Any other thoughts? I think,

256 00:20:57.250 00:21:11.440 Samuel Roberts: they weren’t… so they didn’t realize that it was in there. I guess we didn’t let them know last night, but they saw it today, so hopefully they’ll get some testing in. I got some deployment scripts that I was working on yesterday, but even without that, getting it into prod shouldn’t be crazy today.

257 00:21:11.710 00:21:14.040 Samuel Roberts: Because we’re basically creating prod.

258 00:21:14.140 00:21:17.439 Samuel Roberts: From staging, once it’s in there. So, yeah.

259 00:21:17.440 00:21:31.729 Uttam Kumaran: On my end, I’m trying not to send messages after 6pm to any clients, because it just sends a bad precedent. I don’t know, Claire, it’s like, what you think about that, but, like, I have a really bad habit of…

260 00:21:31.850 00:21:36.330 Uttam Kumaran: Just being online for clients, whenever they happen to be online.

261 00:21:36.550 00:21:42.629 Uttam Kumaran: And I’m trying to shake that, because it sets the whole team up for failure. So, like…

262 00:21:43.470 00:21:54.820 Uttam Kumaran: what I would ask for is, like, to think through, like, by 5pm, like, if I can get an update out to them. I kind of really don’t want to be talking to them at, like, 8 o’clock at night.

263 00:21:54.820 00:21:59.000 Samuel Roberts: No, that’s definitely, that’s definitely… It’s probably working, but, like, I really don’t want to, because…

264 00:21:59.000 00:22:02.710 Uttam Kumaran: Because what’ll happen is they’ll ask me for something, and then no one else will be on.

265 00:22:02.710 00:22:03.709 Samuel Roberts: And I’m like…

266 00:22:03.710 00:22:10.309 Uttam Kumaran: in a jam, right? So I’m just trying to, like, not do that. Some other clients really will abuse us.

267 00:22:10.430 00:22:13.380 Samuel Roberts: Because they’re, like, kind of schizo with work, some people.

268 00:22:13.380 00:22:19.160 Uttam Kumaran: And they just, like, are very demanding, and I really just want to, like, set some boundaries.

269 00:22:19.160 00:22:20.120 Samuel Roberts: I think that makes sense.

270 00:22:20.120 00:22:25.319 Uttam Kumaran: We can always, like, do that in an emergency, but if we’re always there every day, like, 8, 9 o’clock.

271 00:22:25.670 00:22:30.619 Uttam Kumaran: Like, that’s bad consult… that’s, like, bad consulting 101. Right, right. Not from my world.

272 00:22:30.620 00:22:37.650 Samuel Roberts: Yeah, no, I think let’s… let’s keep it up with the end of… end of day updates, by, like, 5… 5.30 here.

273 00:22:37.650 00:22:42.069 Uttam Kumaran: I think it starts to send, like, a note also, like, around, like, 4, being like.

274 00:22:42.610 00:22:46.740 Uttam Kumaran: Give me something to share, you know? Or I could send a reminder or something, yeah.

275 00:22:47.060 00:22:48.110 Samuel Roberts: Okay, cool.

276 00:22:49.170 00:22:56.439 Samuel Roberts: But yeah, I think once we get this out, they’ll be able to actually use it and get it out to their team, and I think we’re still waiting on the…

277 00:22:56.870 00:23:02.269 Samuel Roberts: CSV with the… keys and stuff, so I’ll maybe ping them about that if they want.

278 00:23:02.270 00:23:02.730 Pranav Narahari: Right.

279 00:23:02.730 00:23:09.659 Samuel Roberts: That’s really just, like, a helpful thing, but it’s not really critical for them. They could do it manually, but…

280 00:23:10.650 00:23:19.859 Pranav Narahari: Quick question on, the scheduled Slack messages errors. Should I be making, like, bug fix branches off of staging for that, or should I still go through dev to staging?

281 00:23:22.760 00:23:25.580 Samuel Roberts: Dev was really just for, like, sharing, is my thought.

282 00:23:26.210 00:23:36.190 Samuel Roberts: that we can, like, get a feature up and have a PR environment that’s not just gonna go right into staging, necessarily. I think you’re fine for that to go right to staging for the fixes.

283 00:23:36.190 00:23:36.910 Pranav Narahari: Okay.

284 00:23:37.850 00:23:42.599 Uttam Kumaran: Yeah, the way most places work, I don’t know, Sam, how you did it, like, most places will do, like.

285 00:23:43.150 00:23:48.849 Uttam Kumaran: Once a… they’ll do… they’ll… a lot of places will try to do, like, daily releases, but for the most part, typically, you’re, like.

286 00:23:49.080 00:23:55.129 Uttam Kumaran: pushing everything to one environment and cutting off, like, a release branch, like, on Thursdays with, like.

287 00:23:55.130 00:23:55.560 Samuel Roberts: Yeah.

288 00:23:55.560 00:23:57.600 Uttam Kumaran: features, right? Yep.

289 00:23:57.600 00:23:59.100 Samuel Roberts: Yeah, that’s what I’m thinking from staging.

290 00:23:59.100 00:24:03.890 Uttam Kumaran: These days, it’s not at… most people are actually just, like, doing full deploys, like…

291 00:24:04.520 00:24:09.719 Uttam Kumaran: like, Stripe, for example, they’re very famous for just, like, they deploy, like, tons of times per day.

292 00:24:09.830 00:24:17.960 Uttam Kumaran: Right. But we don’t necessarily have to do that. Again, just want to be conscious of setting a bad precedent on, like.

293 00:24:17.960 00:24:18.510 Samuel Roberts: Yeah.

294 00:24:18.510 00:24:19.890 Uttam Kumaran: speed and quality.

295 00:24:20.220 00:24:23.880 Uttam Kumaran: Because, as you can tell, like, if it doesn’t get shipped well.

296 00:24:24.010 00:24:31.410 Uttam Kumaran: then it’s gonna come back to us. Right. So another ask I would have is, like, for you guys to start thinking about, like, how to put in, like, unit testing.

297 00:24:31.780 00:24:32.140 Samuel Roberts: -

298 00:24:32.140 00:24:37.600 Uttam Kumaran: And a broader, like, test suite. Otherwise, like, we’re gonna have to do a lot more, like.

299 00:24:38.040 00:24:41.540 Uttam Kumaran: visual QA or, like, VQA more often.

300 00:24:41.730 00:24:42.530 Samuel Roberts: Definitely.

301 00:24:44.490 00:24:56.760 Samuel Roberts: Yeah, I was thinking about that with some of the deployment stuff, like testing and making sure that everything is still, you know, unit tests and smoke tests and whatever else we need to get in place, especially for, like, merging into staging and

302 00:24:56.920 00:25:03.340 Samuel Roberts: And obviously production, so… We can maybe discuss some of that later, too, now that we have time.

303 00:25:04.280 00:25:05.879 Samuel Roberts: I have some thoughts there.

304 00:25:06.310 00:25:06.990 Uttam Kumaran: Cool.

305 00:25:07.730 00:25:10.810 Samuel Roberts: Alright, anything else there?

306 00:25:13.940 00:25:15.170 Samuel Roberts: Cool, alright.

307 00:25:15.510 00:25:17.109 Samuel Roberts: I guess let’s talk internal.

308 00:25:17.600 00:25:19.139 Gabriel Lam: Yeah,

309 00:25:19.840 00:25:37.660 Gabriel Lam: I guess I’ll start with the partnership stuff, just for UTAM. I have the demos, I put it on the sales partnership link, but was hoping that you or Holly or anyone would just have a quick eye on it before I put it with the channel, with the, the guys from Contextual.

310 00:25:37.830 00:25:40.339 Gabriel Lam: Aside from that, for the rest of the team.

311 00:25:40.420 00:25:58.349 Gabriel Lam: last week was mainly focused on connecting Cursor to Notion and Google Workspace, so that… the baseline, to recap, is that it is connected, so this week, I’m just hoping to refine that with the ops, team, and just, like, having, you know, regular meetings to see

312 00:25:58.620 00:26:04.529 Gabriel Lam: what it is exactly that they need. In terms of a roadmap, I’m also curious, Utam, if there are things

313 00:26:05.180 00:26:16.649 Gabriel Lam: that you would like to prioritize. It seems like my main goal has really just been meeting up with the different leads, and just making sure people are able to use the platform and cursor, and…

314 00:26:17.180 00:26:18.500 Gabriel Lam: Essentially.

315 00:26:18.770 00:26:24.260 Gabriel Lam: you know, whatever pain points people have to address that. But if there’s a sort of more proactive

316 00:26:24.550 00:26:29.790 Gabriel Lam: like, priority that you guys are envisioning, I’m happy to also start on that, too.

317 00:26:30.870 00:26:37.570 Uttam Kumaran: Okay, cool, yeah, like, maybe let’s talk in the… let’s talk like me, you, and, like, Shashu can talk.

318 00:26:37.750 00:26:43.110 Uttam Kumaran: Yeah, so I’ll take a look at those. Where is the code for the Google Drive MCP?

319 00:26:43.490 00:26:44.809 Uttam Kumaran: It’s, like, in platform.

320 00:26:45.100 00:26:46.259 Uttam Kumaran: Is that a branch?

321 00:26:46.700 00:26:50.599 Gabriel Lam: It’s in… where did I put it?

322 00:26:53.040 00:26:58.499 Gabriel Lam: It’s in the playbook on onboarding called Set Up Google Workspace MCP.

323 00:26:58.500 00:27:01.149 Samuel Roberts: Yeah, I think it’s just another… another repo.

324 00:27:01.150 00:27:02.590 Uttam Kumaran: But we’re not hope… oh, but you’re.

325 00:27:02.590 00:27:07.790 Gabriel Lam: No, no. Well, I’m deploying someone else’s. I’ll… I’ll link that.

326 00:27:07.790 00:27:09.510 Uttam Kumaran: Oh, I see, I see.

327 00:27:10.060 00:27:18.560 Uttam Kumaran: One thing that… one thing that I want to try to do is just continue to, like, I think the game is gonna be to leverage platform as just a big monorepo.

328 00:27:18.560 00:27:19.180 Samuel Roberts: Yeah.

329 00:27:19.440 00:27:27.869 Uttam Kumaran: And because I think we’re gonna end up with, like, a couple of different Slack assistants, I’m gonna set up some more MCPs myself.

330 00:27:28.190 00:27:35.100 Uttam Kumaran: So, I think, basically, I can also help you do that, Gabe. Like, if you send me, like.

331 00:27:35.260 00:27:37.790 Uttam Kumaran: I think you… I saw the railway deployment.

332 00:27:38.230 00:27:41.039 Uttam Kumaran: I can also just… we can just fork that and…

333 00:27:41.860 00:27:45.930 Uttam Kumaran: Or I can just basically copy it into our platform repo and help you deploy that, too.

334 00:27:46.210 00:27:46.600 Gabriel Lam: Okay.

335 00:27:46.600 00:27:54.350 Uttam Kumaran: So, ideally, like, any sort of custom MCPs that we have, I’d like us to just deploy those within our platform repo.

336 00:27:55.060 00:27:56.020 Uttam Kumaran: Okay.

337 00:27:56.590 00:28:01.410 Uttam Kumaran: And then the other thing on, on, like, the playbook and vault,

338 00:28:02.640 00:28:05.950 Uttam Kumaran: One thing that I’m sort of trying to think about is, like.

339 00:28:06.150 00:28:10.840 Uttam Kumaran: There’s two kind of resources that I feel like we need to think about where they live. One is, like.

340 00:28:11.030 00:28:14.550 Uttam Kumaran: sometimes I come across, like, articles, or, like.

341 00:28:15.030 00:28:23.020 Uttam Kumaran: other playbooks, like, how to do something the best way, and it’s like, I kinda wanna just shove that somewhere, so I just want to know, like.

342 00:28:23.230 00:28:24.930 Uttam Kumaran: If that should just end up, like.

343 00:28:25.210 00:28:28.339 Uttam Kumaran: in playbooks, like, under a resource. For example, it could be, like.

344 00:28:28.340 00:28:29.180 Gabriel Lam: Yeah.

345 00:28:29.180 00:28:30.569 Uttam Kumaran: Could be a good, like, book.

346 00:28:31.050 00:28:34.359 Uttam Kumaran: Or it could be, like, a long article that’s, like.

347 00:28:34.940 00:28:39.049 Uttam Kumaran: sort of a canonical article on, like, X topic, right? So, like.

348 00:28:39.180 00:28:40.890 Uttam Kumaran: Being able to put that somewhere.

349 00:28:41.240 00:28:43.329 Gabriel Lam: And then the other thing is sort of, like…

350 00:28:43.330 00:28:47.430 Uttam Kumaran: Yeah, the other thing that’s sort of, like, helpful… Like…

351 00:28:48.400 00:28:53.049 Uttam Kumaran: tools. Like, one thing I was thinking about was…

352 00:28:53.240 00:28:59.700 Uttam Kumaran: Creating a tool that helps people just, like, convert Like, articles or transcripts?

353 00:29:00.110 00:29:03.889 Uttam Kumaran: Into, like, quick, like, basically, like, mini-podcasts.

354 00:29:04.450 00:29:08.640 Uttam Kumaran: Because sometimes I, like, I want to rewatch a meeting.

355 00:29:08.920 00:29:14.680 Uttam Kumaran: But I’m not interested in watch… I can’t, like, I just humanly cannot watch it, like, a couple hour-long meetings.

356 00:29:14.680 00:29:15.320 Samuel Roberts: Right.

357 00:29:15.320 00:29:17.600 Uttam Kumaran: like, I can use AI to give me, like, a…

358 00:29:17.890 00:29:21.170 Uttam Kumaran: 15-minute podcast I can listen to on 2X.

359 00:29:21.410 00:29:33.240 Uttam Kumaran: Right. So I’m thinking about these tools that I can do. I feel like… I guess that may end up in platform, but I feel like I’m starting to put more brain power to this, so I’ll just have questions like that that

360 00:29:33.460 00:29:38.410 Uttam Kumaran: I kind of, Gabe, want you to be a little bit of the filter to decide as, like.

361 00:29:38.690 00:29:48.079 Uttam Kumaran: I’ll… you can kind of consider me, like, the super business user, and so a lot of my questions, I’m sure people will have as they start leveraging the platform, so…

362 00:29:48.740 00:29:49.550 Gabriel Lam: Awesome, okay.

363 00:29:49.550 00:29:59.629 Clarence Stone: Speaking of questions, how are y’all keeping your repos up to date? Like, are you guys doing a git pull every time you start using this, or is there a trick here that I’m missing?

364 00:30:01.950 00:30:04.070 Samuel Roberts: On the, like, locally, yeah.

365 00:30:04.070 00:30:04.710 Clarence Stone: Yeah.

366 00:30:04.710 00:30:11.810 Uttam Kumaran: Yeah, so locally, as long as you have the main branch checked out, yeah, I tend to just pull…

367 00:30:12.050 00:30:15.449 Uttam Kumaran: when I… when I open it, sort of just habitually…

368 00:30:15.720 00:30:16.200 Clarence Stone: Okay.

369 00:30:16.930 00:30:18.710 Gabriel Lam: I also have a desktop.

370 00:30:19.400 00:30:20.139 Uttam Kumaran: Yeah, yeah.

371 00:30:20.140 00:30:20.850 Gabriel Lam: option.

372 00:30:21.630 00:30:24.519 Clarence Stone: I’m… I’m thinking about just making it executable.

373 00:30:25.510 00:30:30.969 Uttam Kumaran: Yeah, like, I think you can run a… you can basically have a cron that, like, pulls from main every X amount of time.

374 00:30:31.110 00:30:32.070 Uttam Kumaran: Yeah. Yeah.

375 00:30:34.030 00:30:52.750 Samuel Roberts: One thought on all that was, if you’re gonna fork that MCP, I… and we move it into the repo, there’s two things we could do about that. One is just putting all the files in and adding it to our repo, but then we might lose if there’s any updates or anything like that that we need to, like, hatch in.

376 00:30:52.750 00:30:56.430 Uttam Kumaran: Yeah, I just… I just was gonna check if that guy ever, like, even hasn’t been updated.

377 00:30:56.430 00:30:59.580 Samuel Roberts: That’s what I was gonna say, see if there’s anything there. The other thing is, there are…

378 00:30:59.580 00:31:00.750 Uttam Kumaran: the alternative.

379 00:31:00.750 00:31:04.540 Samuel Roberts: submodules, and so you can have, like, a Git repo in a Git repo.

380 00:31:04.540 00:31:06.120 Uttam Kumaran: Oh, really? I haven’t done that much.

381 00:31:06.120 00:31:10.010 Samuel Roberts: so I don’t know too much about, like, the ergonomics of that, but I know there’s a.

382 00:31:10.010 00:31:17.069 Uttam Kumaran: Can you just Slack… can you just Slack me that word, or put that in the channel? Yeah, I’ll… Yeah, yeah. There’s a couple repos that I need to do this with.

383 00:31:17.070 00:31:26.200 Samuel Roberts: Yeah, I was actually just watching a video where I was breaking down, like, the benefits of monorepos versus multi-repos, and he was starting to talk about it, and I don’t know that much about submodules, so it could be worse.

384 00:31:26.440 00:31:31.459 Samuel Roberts: a dive into, because I think you’re right, having everything in one place will be more beneficial.

385 00:31:32.330 00:31:36.999 Uttam Kumaran: And then, Gabe, I’m gonna ship the Supabase Vault’s transcripts thing today.

386 00:31:37.370 00:31:39.849 Uttam Kumaran: Yup. Sounds like that should all get figured out.

387 00:31:40.530 00:31:42.909 Samuel Roberts: Is that the transcripts to the repo?

388 00:31:43.240 00:31:43.950 Uttam Kumaran: Yeah.

389 00:31:44.200 00:31:45.370 Samuel Roberts: Okay, yeah, I think…

390 00:31:45.370 00:31:46.420 Uttam Kumaran: How you told me to do it.

391 00:31:46.830 00:31:51.889 Samuel Roberts: Ryan had asked me something about that, and I wasn’t sure if it was the same thing you were asking about yesterday, so I wasn’t…

392 00:31:52.680 00:31:58.249 Samuel Roberts: I don’t know. You may want to sync with him and make sure you guys aren’t duplicating effort right now.

393 00:31:58.820 00:31:59.510 Uttam Kumaran: Okay, okay.

394 00:32:01.550 00:32:05.999 Pranav Narahari: On that podcast thing that you were mentioning, have you guys tried Notebook LM?

395 00:32:06.760 00:32:09.609 Uttam Kumaran: Yeah, I use it. It’s a little bit like…

396 00:32:10.080 00:32:15.699 Uttam Kumaran: I don’t know if you tried it, it’s good, but it’s… now, the kind of the gimmicky shit has kind of worn off for me.

397 00:32:15.700 00:32:16.660 Pranav Narahari: Yeah, yeah.

398 00:32:16.660 00:32:21.079 Uttam Kumaran: Like, I, I’m truly, like, very, very low on time.

399 00:32:21.310 00:32:28.129 Uttam Kumaran: So I… everything needs to be, like, incredibly dense to X, like, otherwise it’s very hard.

400 00:32:28.310 00:32:30.439 Uttam Kumaran: So, like, it’s… that one is good.

401 00:32:30.560 00:32:32.850 Uttam Kumaran: But also, it’s, like, not steerable.

402 00:32:33.060 00:32:34.030 Pranav Narahari: Yeah, yeah.

403 00:32:34.030 00:32:36.679 Uttam Kumaran: you sort of, like, throw a bunch of PDFs in, and then…

404 00:32:36.810 00:32:48.170 Uttam Kumaran: it’s, like, it’s meant to be, like, a nice, like… I think it’s meant to, like, show the power of AI, but, like, once you kind of… once that wears off, I’m kind of like, yo, am I… is this, like, do they miss stuff? Like, how do I know?

405 00:32:48.170 00:32:49.159 Samuel Roberts: Right, right.

406 00:32:49.160 00:32:59.169 Uttam Kumaran: So, what we’re probably better off doing is have, like, a Brainforge standard, like, summarizer, and then I’m gonna pass that to…

407 00:32:59.260 00:33:10.409 Uttam Kumaran: an open source, like, TTS that will just do the podcast. So what I’ll do is I’ll take a bunch of docs in a way that we expect, I’ll have it summarized.

408 00:33:10.560 00:33:16.449 Uttam Kumaran: Like, you know, so… What to not… what to never skip over, what to skip over.

409 00:33:17.340 00:33:24.180 Uttam Kumaran: For, you know, for example, like, a lot of our summaries will be like, so-and-so went to a birthday this weekend, like, that’s, like, not relevant at all.

410 00:33:24.180 00:33:25.210 Samuel Roberts: Right, right.

411 00:33:25.210 00:33:25.820 Uttam Kumaran: Right?

412 00:33:26.100 00:33:26.620 Pranav Narahari: Yeah.

413 00:33:26.620 00:33:28.810 Uttam Kumaran: Instead, I’m much more interested in, like.

414 00:33:29.240 00:33:33.110 Uttam Kumaran: Like, okay, and… but you can even take it one step further.

415 00:33:33.110 00:33:33.880 Samuel Roberts: like…

416 00:33:33.880 00:33:48.979 Uttam Kumaran: Eventually, in the platform, we’re gonna have personas, like, around each employee, and what’s gonna happen is, like, it will be more and more personalized towards you. For example, my podcast may not be the podcast that’s most important to you.

417 00:33:49.050 00:34:02.369 Uttam Kumaran: Right? For example, like, for me, I’m less interested in the details, I’m more like… my prompt would be, hey, you’re the CEO of this company, like, summarize this in a way where the key, like, risks

418 00:34:02.660 00:34:13.499 Uttam Kumaran: And, like, the key communication points are, like, summarized, right? But for someone who’s, like, on engineering, it actually may be less about that, and maybe actually, like, explaining, like, what it is to plan.

419 00:34:13.620 00:34:27.929 Uttam Kumaran: And so, that’s… we’re gonna, like, probably later this year, think more about, like, how the platform is, sort of morphs to each role, a little bit more heavily. But for now, yeah, I’m gonna just have, like, a Brainforge

420 00:34:29.590 00:34:36.469 Uttam Kumaran: podcast prompt that will kind of write it out, like the Notebook LM, like maybe two people discussing.

421 00:34:36.670 00:34:39.189 Uttam Kumaran: Because I feel like that format is kind of nice, but…

422 00:34:39.190 00:34:40.169 Samuel Roberts: It is a nice format, you know.

423 00:34:40.170 00:34:43.109 Uttam Kumaran: It will be… it’ll be really steerable, and…

424 00:34:43.440 00:34:45.790 Uttam Kumaran: Yeah, I’m gonna try to just, like…

425 00:34:46.480 00:34:55.079 Uttam Kumaran: One thing that’s, like, really tough in this business is that, like, there’s a lot of meetings going on, and it’s hard to understand

426 00:34:55.219 00:34:58.929 Uttam Kumaran: where I need to be in. So my typical ask for everybody here is, like.

427 00:34:59.080 00:35:11.840 Uttam Kumaran: yo, if you feel like there’s a place for me to help prioritize, or there’s a red flag, call me in. But that’s also something that it’s, like, it’s kind of dependent on our team to do, versus, like, if I can sort of have something

428 00:35:12.130 00:35:23.999 Uttam Kumaran: find those and surface them up, like, it makes it easier for me to kind of be helpful, even in areas where people may not know that, like, they can loop me in. So that’s kind of, like, the direction I’m heading.

429 00:35:26.820 00:35:28.599 Samuel Roberts: Gotcha. Yeah, that makes sense.

430 00:35:29.590 00:35:42.310 Pranav Narahari: Yeah, honestly, that feature for, like, podcasting I never really used, but, for the reasons, kind of like you said. But then I also saw something, like, there’s, like, an interactive, like, discussion you could have, too, so I don’t know…

431 00:35:42.390 00:35:52.530 Pranav Narahari: it’s just, like, kind of ideas for us to think about, like, what we want to have in the platform, because sometimes if you can ask a question, that could…

432 00:35:53.100 00:36:09.249 Pranav Narahari: like, if you have a question, you don’t need to wait for it to be, like, discussed in the podcast. You can just, like, ask it straight up in the beginning, and then that can then rabbit hole into, like, a deeper discussion. And maybe you gain more information faster about the project that way, but…

433 00:36:09.250 00:36:10.619 Uttam Kumaran: Yeah, makes sense.

434 00:36:10.620 00:36:11.270 Pranav Narahari: Yeah.

435 00:36:18.450 00:36:21.330 Samuel Roberts: Any other… Thoughts on any of that stuff?

436 00:36:25.280 00:36:26.910 Samuel Roberts: Alright, sweet.

437 00:36:27.240 00:36:32.160 Samuel Roberts: Thank you all. Yeah, no, definitely set up that time later.

438 00:36:32.450 00:36:36.910 Samuel Roberts: for us to chat, through the Lilo stuff, and then…

439 00:36:37.270 00:36:40.830 Samuel Roberts: Yeah, if anyone needs me, definitely let me know.

440 00:36:41.220 00:36:43.420 Samuel Roberts: Otherwise, have a good day.

441 00:36:45.120 00:36:45.780 Pranav Narahari: See you guys.

442 00:36:45.780 00:36:47.189 Samuel Roberts: Alright, thanks everyone.