Meeting Title: Brainforge x Lilo Task Estimates Date: 2026-02-17 Meeting participants: Samuel Roberts, Casie Aviles, Pranav Narahari


WEBVTT

1 00:00:53.210 00:00:54.170 Pranav Narahari: Hey guys.

2 00:00:57.560 00:00:57.940 Samuel Roberts: Hey.

3 00:01:01.900 00:01:04.319 Pranav Narahari: Yeah, so Bobby kind of got back to me.

4 00:01:04.489 00:01:12.070 Pranav Narahari: about 30 minutes ago or something, I got additional clarity on, A few things.

5 00:01:12.530 00:01:19.209 Pranav Narahari: But, yeah, there’s basically 5 line items that I’ve highlighted in that Stitch backlog.

6 00:01:19.350 00:01:24.639 Pranav Narahari: sheet. I don’t know if I’ve already sent that to y’all, or if you guys have access to it. I can.

7 00:01:24.640 00:01:25.380 Samuel Roberts: Where is it?

8 00:01:26.450 00:01:30.609 Pranav Narahari: It should be… I think I forwarded it into our channel, but…

9 00:01:30.610 00:01:31.260 Samuel Roberts: Okay.

10 00:01:31.260 00:01:32.390 Pranav Narahari: I’m not sure if it was…

11 00:01:32.890 00:01:35.980 Samuel Roberts: I just wasn’t sure if it was Notion or Google or what, I couldn’t remember.

12 00:01:36.270 00:01:37.380 Pranav Narahari: Yeah, it was a ghoul shoot.

13 00:01:37.650 00:01:40.590 Samuel Roberts: Yeah, that’s right, okay, yeah, alright, I remember this now, thank you.

14 00:01:45.370 00:01:46.240 Samuel Roberts: Okay.

15 00:01:48.290 00:01:50.669 Samuel Roberts: So, what are we looking at?

16 00:01:51.570 00:01:56.720 Pranav Narahari: Yep, so there’s a few different line items here. Sorry, can’t talk.

17 00:01:56.970 00:02:05.320 Pranav Narahari: Yeah, the data warehouse, forecasting updates, meta dashboard data review, prompt management, and then…

18 00:02:06.370 00:02:08.479 Pranav Narahari: Yeah, add knowledge base per brand.

19 00:02:08.919 00:02:13.259 Pranav Narahari: So… A few of these things, like, I feel…

20 00:02:13.900 00:02:22.770 Pranav Narahari: I have, like, rough estimates for the data warehouse. So in the description, you’ll see Move to AirByte, that’s what they wrote. Yeah.

21 00:02:22.940 00:02:28.940 Pranav Narahari: we don’t really need additional description here, even though it’s, like, no description. I feel like we have enough context.

22 00:02:30.840 00:02:34.050 Pranav Narahari: Just so we can, like, sink a little bit more on that.

23 00:02:34.420 00:02:40.690 Pranav Narahari: why they said move to AirByte is, and I think I kind of talked about the polyatomic airbite situation, right?

24 00:02:42.130 00:02:42.850 Samuel Roberts: Yeah.

25 00:02:42.850 00:02:48.600 Pranav Narahari: I did, right? Okay. So… After discussing with Bobby today, he said.

26 00:02:49.180 00:02:52.089 Pranav Narahari: just going forward, let’s just use Airbyte for everything else.

27 00:02:52.290 00:02:57.530 Pranav Narahari: So, he hasn’t really… Specifically said, like, whether we can…

28 00:02:57.860 00:03:06.020 Pranav Narahari: we need to, in the future, maybe down the line, come back to this Shopify integration with Polytomic, but for right now, we can…

29 00:03:06.210 00:03:11.130 Pranav Narahari: just leave Shopify as is with Polytomic, and…

30 00:03:11.240 00:03:14.870 Pranav Narahari: just kind of use AirByte for all the future connections, so…

31 00:03:15.160 00:03:17.089 Samuel Roberts: Basically, we’ll have to just, yeah.

32 00:03:17.280 00:03:20.060 Pranav Narahari: use Meta, Google, and

33 00:03:20.300 00:03:25.019 Pranav Narahari: whatever else is in the future, using AirByte, and then connecting that to Mother Duck.

34 00:03:25.700 00:03:31.269 Pranav Narahari: So, I’ve had that estimated at… let me just read my sheet again.

35 00:03:33.450 00:03:41.180 Pranav Narahari: I think we can probably do that in, like… 40 hours?

36 00:03:41.400 00:03:45.520 Pranav Narahari: That’s what I’m thinking. Basically, it’s like… one week.

37 00:03:47.020 00:03:53.039 Pranav Narahari: I’m not sure if you guys think we should have longer on this, just considering how long, like, Shopify took.

38 00:03:54.240 00:04:00.520 Samuel Roberts: Yeah, I would also wonder, like, when we say data warehouse move to air byte, we mean syncing…

39 00:04:00.660 00:04:11.090 Samuel Roberts: All 70 brands, or… You know, like, are we talking all… all the sources for all brands?

40 00:04:11.770 00:04:13.040 Pranav Narahari: Yeah, yeah.

41 00:04:16.630 00:04:25.870 Samuel Roberts: Yeah, alright, I mean, I think if there’s no… Major hiccups there.

42 00:04:28.950 00:04:33.900 Samuel Roberts: That a… yeah, so it’s like a solid week of work, but that’s not really a week, is it? So…

43 00:04:34.230 00:04:37.960 Samuel Roberts: Like… 40 hours might be fine.

44 00:04:41.130 00:04:46.789 Pranav Narahari: We can also come back to this, like, that’s what I’m thinking, and then I think, thinking about, like, the total amount.

45 00:04:47.210 00:04:48.210 Pranav Narahari: And…

46 00:04:48.470 00:05:04.400 Pranav Narahari: that might make more sense, like, okay, because this data warehouse component of things is just, like, a part of the entire puzzle. I think we also can kind of estimate, like, okay, for an app of this complexity, for, like, this set of features, like, can we get this done within a month, within, like…

47 00:05:04.420 00:05:12.699 Pranav Narahari: 6 weeks. Yeah, yeah. So that’s another way we can think about it. So, okay, I’m glad, like, 40 hours sounds okay. I’m thinking maybe we have to increase it, but…

48 00:05:13.200 00:05:14.650 Pranav Narahari: We can come back to it.

49 00:05:15.000 00:05:22.490 Samuel Roberts: Yeah, I mean… There’s still some question marks around how they auth, right?

50 00:05:23.500 00:05:24.380 Pranav Narahari: Yeah.

51 00:05:24.380 00:05:32.160 Samuel Roberts: because we had… Shopify was up, but we’re not worrying about that one. Meta… they… We’re going to off?

52 00:05:32.160 00:05:34.780 Pranav Narahari: But that was polyatomic, so now it’s different. Okay, never mind.

53 00:05:34.960 00:05:37.089 Samuel Roberts: Sorry, I’m just thinking out loud. Okay.

54 00:05:37.360 00:05:39.019 Samuel Roberts: That’s fine, yeah, stick with that for now.

55 00:05:39.370 00:05:47.820 Pranav Narahari: Okay, cool. And then within here, yeah, there’s the forecasting feedback description in the description for forecasting updates.

56 00:05:50.210 00:05:53.400 Pranav Narahari: they’re not really just, like, small updates. It looks like…

57 00:05:54.090 00:06:08.339 Pranav Narahari: they are… this is kind of deeply integrated with the data warehouse stuff as well, because we need to be pulling in meta and Google stuff. Once we have it in Mother Duck, bringing in that info is actually pretty easy, I would say.

58 00:06:08.520 00:06:13.089 Samuel Roberts: Yeah, just, like, a little bit of modeling and some math, like… Yeah.

59 00:06:13.480 00:06:15.690 Samuel Roberts: If even that. Okay.

60 00:06:16.820 00:06:17.400 Pranav Narahari: Yep.

61 00:06:17.790 00:06:20.230 Samuel Roberts: So that’s definitely, yeah, it’s dependent on that, okay, yeah.

62 00:06:20.350 00:06:28.909 Pranav Narahari: Yeah, and then at the bottom, it’s just, like, some calculation stuff, which is… looks complex, but I think it’s really just, like, we just need to apply that logic into…

63 00:06:28.910 00:06:31.319 Samuel Roberts: Yeah, I don’t think it’d be crazy to do that.

64 00:06:31.320 00:06:32.790 Pranav Narahari: Pretty simple, yeah.

65 00:06:32.900 00:06:38.730 Pranav Narahari: So, on this one… Well, if we think about it.

66 00:06:38.910 00:06:42.760 Pranav Narahari: We can think of data coming into the data warehouse as…

67 00:06:42.960 00:06:49.099 Pranav Narahari: The first task we just talked about, and then the forecasting update, which is basically querying that data.

68 00:06:49.490 00:06:55.880 Pranav Narahari: From Mother Duck into… The forecast tool to be, like, another task that’s…

69 00:06:56.070 00:07:03.550 Pranav Narahari: as part of this forecasting updates task, which, you know, creating that query for Shopify did take a minute.

70 00:07:03.730 00:07:18.630 Pranav Narahari: We can… I assume it’s not gonna be as difficult, but since there’s 3 other connections, we can still assume maybe those 3 will take just as long, or if maybe a little bit longer than it took to get that one Shopify query.

71 00:07:18.630 00:07:19.300 Samuel Roberts: Yeah.

72 00:07:20.010 00:07:25.130 Pranav Narahari: So… This is another thing which I would… I think probably, like.

73 00:07:26.020 00:07:29.720 Pranav Narahari: Another, like, 30 to 40 hours?

74 00:07:30.550 00:07:37.230 Pranav Narahari: to update, does that sound like the right ballpark, Sam?

75 00:07:37.230 00:07:40.370 Samuel Roberts: Sorry, let me just look a little closer at this feedback.

76 00:07:48.090 00:07:54.080 Samuel Roberts: Yeah… Yeah, maybe 30’s fine.

77 00:08:00.000 00:08:01.589 Pranav Narahari: Okay, I think that… that sounds.

78 00:08:01.590 00:08:04.039 Samuel Roberts: Yeah, I would go with that for now. I mean, that might…

79 00:08:04.840 00:08:10.809 Samuel Roberts: change a little bit, but I think once the data’s there, it’s not crazy, so if we’re not including any of that, you know.

80 00:08:12.110 00:08:16.020 Samuel Roberts: If everything’s in the warehouse and this is a separate one, then yeah, I would say that’s fine.

81 00:08:16.480 00:08:18.100 Samuel Roberts: Yeah. Excuse me.

82 00:08:18.100 00:08:23.800 Pranav Narahari: Also, I’m gonna pause here, because one thing I wanna mention about…

83 00:08:24.420 00:08:28.370 Pranav Narahari: phase one that I’m kind of just, like, thinking about is…

84 00:08:28.750 00:08:34.419 Pranav Narahari: I mean, I don’t know if we call it Phase 1, but just basically for January up until, like, right now, is…

85 00:08:34.750 00:08:51.180 Pranav Narahari: we, like, pushed extremely quickly, for Lilo, and, basically I ran the numbers to just see, like, okay, how many… because how our deal works with Lilo, right? Just kind of high level, I don’t need to get into, like, the super details unless you guys want me to.

86 00:08:51.180 00:08:55.950 Pranav Narahari: Is… we’re all… we just pay a fixed monthly amount to them.

87 00:08:55.980 00:09:05.740 Pranav Narahari: And… Basically, the hours that… We basically tried to use a number of hours internally so that the client.

88 00:09:05.740 00:09:06.339 Samuel Roberts: Right, right.

89 00:09:06.340 00:09:08.439 Pranav Narahari: expense for us, right? .

90 00:09:08.440 00:09:09.210 Casie Aviles: That’s, yep.

91 00:09:09.210 00:09:16.190 Pranav Narahari: And so, basically, we did spend a lot more hours than… makes…

92 00:09:16.320 00:09:22.350 Pranav Narahari: sense. And I think that’s just because, you know, not because we were working slow at all. I think it was just because, like.

93 00:09:22.520 00:09:32.789 Pranav Narahari: we were just pushing super fast for them. A lot of ad hoc stuff came in that I just kept on saying yes, yes, yes to, and so I don’t want to make that same mistake for us again.

94 00:09:32.790 00:09:33.260 Samuel Roberts: playing.

95 00:09:33.260 00:09:35.330 Pranav Narahari: Yeah, and like, Sam, I feel like…

96 00:09:35.940 00:09:43.520 Pranav Narahari: you were working a lot on Lilo, which I don’t think initially was the… the goal. I think it was more so you would have, like, a high-level, like.

97 00:09:43.520 00:09:47.780 Samuel Roberts: Yeah, I think I kind of felt like I got sucked in after the MCP stuff a little bit, and then…

98 00:09:47.780 00:09:48.340 Pranav Narahari: Yeah.

99 00:09:48.340 00:09:49.630 Samuel Roberts: Get out fast enough.

100 00:09:49.840 00:09:55.700 Pranav Narahari: Yeah, and I think now I’ve cleared that up with, like, Utam, like, he kind of told me, he was like, yeah, I should have let you know, like.

101 00:09:55.990 00:09:58.399 Pranav Narahari: Kind of where we should allocate ours, you know?

102 00:09:58.880 00:10:02.850 Pranav Narahari: And so I also want to keep that in mind, too, like…

103 00:10:03.470 00:10:06.820 Pranav Narahari: When we’re creating, like, these hours estimates, what we can also do

104 00:10:06.950 00:10:08.960 Pranav Narahari: It kind of makes sense to…

105 00:10:09.710 00:10:13.800 Pranav Narahari: Think about them taking a little bit longer. I mean…

106 00:10:14.060 00:10:18.519 Pranav Narahari: It depends on how we think about it. Technically, we don’t, because they’re just ours, you know, and just where…

107 00:10:20.440 00:10:23.539 Pranav Narahari: But what they’re gonna say, basically, is just gonna be…

108 00:10:23.650 00:10:30.899 Pranav Narahari: okay, we want to get things done as soon as possible. That’s what happened last time. I think, initially, we did scope for, okay, Sam, you’re gonna work

109 00:10:31.020 00:10:40.209 Pranav Narahari: reducing the amount of hours. I think we scoped it for, like, totally, out of the three of us, we’re gonna work, like, maybe, like, 40 hours per week, or, like, maybe, like, 50.

110 00:10:40.610 00:10:48.229 Pranav Narahari: And so, I think… basically, just, like, in that deal, we just kind of didn’t…

111 00:10:48.650 00:10:48.980 Samuel Roberts: Yeah.

112 00:10:48.980 00:10:52.500 Pranav Narahari: We plan it correctly, and we just ended up working way more than we were getting.

113 00:10:52.500 00:10:52.940 Samuel Roberts: Right, right.

114 00:10:52.940 00:10:53.740 Pranav Narahari: paid for.

115 00:10:54.550 00:10:55.080 Samuel Roberts: Okay.

116 00:10:55.080 00:10:58.569 Pranav Narahari: That’s just another thing to keep in mind. I don’t know, like, it doesn’t really…

117 00:10:59.840 00:11:02.360 Pranav Narahari: I just wanted y’all to be aware of that as well, since that.

118 00:11:02.360 00:11:07.249 Samuel Roberts: Yeah, totally, totally. Yeah, I’ve been trying to step back a little bit anyway since I,

119 00:11:07.380 00:11:13.470 Samuel Roberts: Yeah, once… once I was, like, not in the middle of anything, I basically had been trying to step back, so I think that’s hopefully…

120 00:11:14.400 00:11:15.969 Samuel Roberts: solves going forward.

121 00:11:16.160 00:11:17.290 Pranav Narahari: Yeah, perfect.

122 00:11:17.840 00:11:28.370 Pranav Narahari: Alright, cool. Okay, so… The Meta Dashboard Data Review.

123 00:11:29.230 00:11:32.690 Pranav Narahari: Let’s see… yeah, a lot of these metrics are slightly off. I think…

124 00:11:33.510 00:11:36.540 Casie Aviles: Yeah, Bobby sent something about this, I believe, yesterday.

125 00:11:36.760 00:11:37.670 Pranav Narahari: Yeah.

126 00:11:38.100 00:11:45.199 Pranav Narahari: Yeah, I think this can probably be, like, a 2-day effort, maybe, like, 16 to 24 hours.

127 00:11:47.240 00:11:48.720 Pranav Narahari: Shouldn’t take too long there.

128 00:11:49.850 00:11:51.480 Samuel Roberts: Yeah, so he thinks it’s just a…

129 00:11:51.590 00:11:53.460 Samuel Roberts: Like, a time zone thing, maybe?

130 00:11:54.070 00:11:55.580 Pranav Narahari: Yeah, something like that.

131 00:11:55.580 00:11:56.320 Samuel Roberts: Okay.

132 00:11:57.720 00:12:00.629 Pranav Narahari: But, and the thing is, here, is that…

133 00:12:01.140 00:12:03.269 Pranav Narahari: This is just, like, a temporary thing.

134 00:12:03.380 00:12:07.950 Pranav Narahari: Because it’s pulling the data from the MCP and not from the data warehouse, where it should be.

135 00:12:08.490 00:12:10.099 Samuel Roberts: Oh, right, right.

136 00:12:10.100 00:12:10.470 Pranav Narahari: Yeah.

137 00:12:10.470 00:12:12.489 Samuel Roberts: Yeah, I would say, yeah, try to keep that

138 00:12:13.050 00:12:15.709 Samuel Roberts: Yeah, 16 hours, I wouldn’t go more than that.

139 00:12:15.930 00:12:17.010 Samuel Roberts: Ideally.

140 00:12:17.260 00:12:21.609 Samuel Roberts: Because it’s just, it’s just, like, effort that’s gonna disappear once we have the warehouse.

141 00:12:22.060 00:12:25.089 Samuel Roberts: So we just kind of need to patch it as quickly as we can.

142 00:12:25.250 00:12:35.939 Pranav Narahari: Yeah, and I’ll let them know that. I’ll just be like… I think what I’ll probably say is that it’s gonna take 24 hours to do if we do use the MCPs, but that’s work that is going to be, like.

143 00:12:36.120 00:12:44.869 Pranav Narahari: basically just replaced by all the data warehouse stuff that we do. And so, is this really, like, a priority to get this done first? And…

144 00:12:44.870 00:12:45.910 Samuel Roberts: Right. Prior to that.

145 00:12:46.520 00:12:54.569 Pranav Narahari: data warehouse stuff? Or can we just, like, do the data warehouse stuff, and then this meta dashboard will then pull in from that… from that data?

146 00:12:54.770 00:12:59.360 Pranav Narahari: from Mother Duck, and then we don’t need to… Worry about it going forward.

147 00:13:01.850 00:13:02.540 Samuel Roberts: Yeah.

148 00:13:02.800 00:13:06.500 Pranav Narahari: Yeah, okay, cool, I’ll tell them that. Okay.

149 00:13:06.660 00:13:08.240 Pranav Narahari: Prompt management.

150 00:13:09.590 00:13:12.160 Pranav Narahari: Yeah, so there’s another dock connected here.

151 00:13:16.240 00:13:23.720 Pranav Narahari: just overall edits. And then in there, there’s this, screenshot for…

152 00:13:24.460 00:13:30.119 Pranav Narahari: yeah, basically this new thing that they’re calling the agent library, which will just, like, replace the prompt library.

153 00:13:30.120 00:13:30.640 Samuel Roberts: Chef.

154 00:13:30.640 00:13:40.180 Pranav Narahari: This is kind of like a larger task, because it’s not just a UI change, it’s going to also allow for them to schedule, responses from the

155 00:13:40.740 00:13:43.109 Pranav Narahari: like, the chat into Slack directly.

156 00:13:45.660 00:13:46.850 Pranav Narahari: So…

157 00:13:46.850 00:13:47.810 Samuel Roberts: Oh, okay.

158 00:13:47.810 00:13:54.010 Pranav Narahari: bit more complex than just, like… Yeah, plus they want, like, tagging and stuff here too, right? Excuse me.

159 00:13:56.200 00:13:59.180 Samuel Roberts: versus branding, or brand-specific?

160 00:14:00.430 00:14:01.570 Pranav Narahari: Let’s see…

161 00:14:02.800 00:14:07.339 Samuel Roberts: Yeah, you need to be able to manage COD skills. I was saying that in the past. Yeah, I see it under… in the…

162 00:14:07.630 00:14:13.230 Samuel Roberts: Well, I don’t know what file you’re in, I just see anonymous Squirrel and Anonymous, like, other things, so I don’t know what’s what here.

163 00:14:13.500 00:14:18.510 Pranav Narahari: Did they mention that in, one of these… Docs?

164 00:14:19.060 00:14:22.980 Samuel Roberts: Yeah, it’s in the Brainforge edits under Problem Management. Oh, I see.

165 00:14:22.980 00:14:24.990 Pranav Narahari: That first line need to be able to manage closets.

166 00:14:25.590 00:14:31.399 Pranav Narahari: Probably or to specific brands. Yeah, okay, so that’s gonna be another, like, probably schema change, or something like that.

167 00:14:31.510 00:14:35.920 Samuel Roberts: Yeah, I think we probably just gonna add some tags, and then… I don’t even know if we need to worry about…

168 00:14:37.070 00:14:37.930 Samuel Roberts: like…

169 00:14:38.800 00:14:46.599 Samuel Roberts: tying… I don’t know. Yeah, we can think that through a little bit, but it shouldn’t be crazy to do that. I think the scheduling is probably a heavier lift.

170 00:14:47.400 00:14:48.180 Pranav Narahari: Okay.

171 00:14:49.860 00:14:50.820 Pranav Narahari: So…

172 00:14:51.920 00:14:59.420 Pranav Narahari: I think everyone’s sick right now. I know, I’m just like, oh my god, what’s going on? Yeah, my kid had a runny nose, and I was like, oh no, it’s gonna… it begins.

173 00:14:59.760 00:15:01.680 Samuel Roberts: So I’m like, sorry, I’m making sounds.

174 00:15:01.990 00:15:06.359 Pranav Narahari: Yeah, no, no, I think I’m doing the same thing, so… all good.

175 00:15:06.950 00:15:10.550 Pranav Narahari: Alright, so… this one, too, I would say…

176 00:15:15.010 00:15:23.239 Pranav Narahari: I’m thinking, like, I kind of don’t like spec- like, getting down to, like, the granular day, because things are always coming up.

177 00:15:23.590 00:15:24.290 Samuel Roberts: Yeah.

178 00:15:25.420 00:15:29.529 Pranav Narahari: I kinda wanna give this a week as well. Like, another 40 hours.

179 00:15:30.000 00:15:31.610 Samuel Roberts: Like…

180 00:15:33.130 00:15:39.500 Samuel Roberts: Yeah, I would say give it the hours, and then in terms of, like, scheduling, like, we get into a Gantt, and we can actually see where it all fits in.

181 00:15:39.930 00:15:45.309 Pranav Narahari: Yeah, so I’m just thinking, like, say if just one person’s working on it, it would take 40 hours, so… Yeah.

182 00:15:45.310 00:15:49.670 Samuel Roberts: Totally, totally, totally. I think, I think that’s… Probably fine.

183 00:15:50.050 00:15:54.590 Samuel Roberts: Maybe a little high, but I’d rather go high than not for the uncertainty of some of that.

184 00:15:54.830 00:16:06.080 Pranav Narahari: Yeah, and that’s the whole thing, too, like, I don’t expect us to, like, fully just take, like, a back seat, you know? Like, we could definitely get things done faster.

185 00:16:06.400 00:16:06.900 Samuel Roberts: Yeah, yeah.

186 00:16:06.900 00:16:07.640 Pranav Narahari: And then…

187 00:16:07.970 00:16:15.199 Pranav Narahari: they’re gonna love that, right? So… it’s just, like, setting us… setting us up for success here.

188 00:16:15.200 00:16:15.880 Samuel Roberts: Yeah.

189 00:16:16.300 00:16:19.119 Pranav Narahari: I’d rather them agree to, like, us, like.

190 00:16:19.260 00:16:25.319 Pranav Narahari: Saying, like, a higher amount, and then we just kill it and, like, get it done in half the time, or 75% of the time.

191 00:16:25.320 00:16:25.920 Samuel Roberts: Yeah.

192 00:16:26.790 00:16:29.600 Samuel Roberts: Yeah, if it rolls out faster to them, they’re gonna be happy anyway, so…

193 00:16:29.600 00:16:34.870 Pranav Narahari: Yeah, so part of this is, like, what are they gonna agree to, right? Like, a week of effort is, like.

194 00:16:36.090 00:16:42.560 Pranav Narahari: it’s kind of hard for them to say no, like, okay, we’ll get something out to y’all in a week, like.

195 00:16:42.560 00:16:46.429 Samuel Roberts: Well, is it a week of effort, or does it get that out to them in a week? Because those are two different…

196 00:16:47.280 00:16:54.400 Pranav Narahari: I think it would be even less than, like, a week, right? Because we’re gonna be probably working more than 40 hours in a week out of the three of us.

197 00:16:54.400 00:16:56.090 Samuel Roberts: Yeah, combined, true, true.

198 00:16:56.310 00:16:56.850 Pranav Narahari: Yeah.

199 00:16:56.850 00:16:57.820 Samuel Roberts: Good point. Okay.

200 00:16:59.480 00:17:04.600 Samuel Roberts: Yeah, alright, that’s fine there, then. And then the knowledge base per brand…

201 00:17:05.230 00:17:19.179 Pranav Narahari: Yeah. So this is something that they briefly mentioned before, but I think this is actually going to be pretty high complexity if we want the… you know, there’s levels to this, right? And the images part of this is like, okay.

202 00:17:19.490 00:17:35.590 Pranav Narahari: if we want to properly create, like, knowledge… image knowledge bases, like, that probably requires, like, embeddings. It’s not as simple as just, like, maybe even with, like, certain small documents, you just add it there as context. But probably for all of this, we’re going to want to create embeddings.

203 00:17:37.450 00:17:39.280 Samuel Roberts: Yeah, I think this one is…

204 00:17:39.280 00:17:41.619 Pranav Narahari: big enough that I think it needs to get, like.

205 00:17:41.620 00:17:45.430 Samuel Roberts: scoped out in, like, what does this mean? You know?

206 00:17:45.750 00:17:46.550 Samuel Roberts: like…

207 00:17:47.900 00:17:53.420 Samuel Roberts: in what level? Like, because I feel like a knowledge base per brand could be a number of… I mean, Sheets…

208 00:17:53.970 00:17:59.269 Samuel Roberts: Okay, documents, images, like, is this all things they want to reference all the time?

209 00:17:59.580 00:18:00.510 Samuel Roberts: Is it…

210 00:18:00.920 00:18:08.040 Samuel Roberts: things they want to call in. Do we need to build a RAG pipeline for every one of these brands? Or not build, you know, but I mean scale it to that.

211 00:18:10.560 00:18:15.920 Pranav Narahari: Yeah, I mean, I think that’s what we should estimate. Like, that’s the complexity, and then…

212 00:18:15.920 00:18:16.470 Samuel Roberts: Okay.

213 00:18:16.470 00:18:23.920 Pranav Narahari: tell us, like, you know, this is, like, way too high of an estimate, then we can talk about, okay, well, we can reduce the complexity here, it will not be as, like…

214 00:18:23.920 00:18:24.840 Samuel Roberts: Yeah, that’s fair.

215 00:18:24.840 00:18:28.170 Pranav Narahari: Yeah, so, like, let’s actually think about it in, like.

216 00:18:28.770 00:18:33.169 Pranav Narahari: how can we ship, like, the best possible product for this? So, like, if we’re…

217 00:18:33.510 00:18:35.780 Pranav Narahari: Creating the separate pipelines per brand.

218 00:18:36.220 00:18:43.720 Pranav Narahari: We want to be able to support all of these different types of, Like…

219 00:18:44.130 00:18:47.440 Pranav Narahari: Documents, and just files.

220 00:18:47.640 00:18:57.139 Pranav Narahari: how can we… how can we do that? I think some of y’all’s, like, ABC work could… could be used as context here, like…

221 00:18:57.620 00:19:01.499 Pranav Narahari: What do y’all think for an estimate on this one? This one is, like, the one I struggled with the most.

222 00:19:06.420 00:19:14.280 Casie Aviles: Yeah, we haven’t… I’m not too sure yet for, like… Images and sheets.

223 00:19:15.200 00:19:18.880 Casie Aviles: Because for sheets, what we…

224 00:19:19.000 00:19:22.299 Casie Aviles: Did was we have, like, a custom…

225 00:19:22.660 00:19:25.300 Casie Aviles: We have, like, a dedicated table.

226 00:19:26.060 00:19:29.369 Casie Aviles: Or, like, database for the sheets.

227 00:19:29.970 00:19:32.630 Samuel Roberts: I see what you’re saying. Yeah, I think…

228 00:19:32.630 00:19:35.859 Casie Aviles: This is the only one we basically did rug on.

229 00:19:37.550 00:19:45.479 Pranav Narahari: So if we were to estimate just the documents, say the knowledge base was just per brand, and it was just documents, what would you estimate that to be?

230 00:19:50.880 00:19:51.760 Casie Aviles: Mmm…

231 00:19:53.540 00:19:59.990 Samuel Roberts: Are we talking about… the idea here is, like, just a markdown file, or, like, a text file that stores brand info.

232 00:20:01.240 00:20:04.469 Samuel Roberts: Or are we talking, like, multiple documents? You know what I mean? Like…

233 00:20:04.620 00:20:07.579 Samuel Roberts: Are they uploading PDFs? Are they uploading…

234 00:20:09.810 00:20:14.390 Pranav Narahari: Yeah, I would say, let’s think of documents as just, like, one file format, just PDF.

235 00:20:14.840 00:20:15.540 Samuel Roberts: Okay.

236 00:20:15.700 00:20:19.710 Pranav Narahari: Or honestly, what might even be easier is just TXT.

237 00:20:20.030 00:20:31.459 Samuel Roberts: That’s what I was thinking. I mean, I think the, like, at the simplest, like, I… part of it is I want to understand what they want from a knowledge base. Without knowing that, I think it’s a little hard, but, like, if it is just, like, they want the…

238 00:20:32.360 00:20:34.740 Samuel Roberts: Chat to have more context.

239 00:20:35.060 00:20:39.759 Samuel Roberts: I think it’s Easy to upload, like, a…

240 00:20:41.170 00:20:45.300 Samuel Roberts: Either a markdown, or have them in an editor in the site, be able to put in

241 00:20:46.320 00:20:48.360 Samuel Roberts: Context we can inject.

242 00:20:48.820 00:20:52.930 Samuel Roberts: If it’s more… than that.

243 00:20:53.590 00:20:57.390 Samuel Roberts: It’s… it’s less… Clear.

244 00:20:57.800 00:21:00.390 Pranav Narahari: So are you saying not even creating embeddings?

245 00:21:01.230 00:21:03.559 Samuel Roberts: Depending on how much they’re trying to put in, yeah.

246 00:21:04.360 00:21:08.200 Samuel Roberts: You know, if it’s… cause if it’s stuff we wanna… if it’s… if it’s something like…

247 00:21:08.420 00:21:11.620 Samuel Roberts: I just don’t know how much data they’re talking here, like, how much knowledge.

248 00:21:12.070 00:21:13.930 Casie Aviles: Yeah, if it’s a lot…

249 00:21:13.930 00:21:14.600 Pranav Narahari: Yeah.

250 00:21:14.600 00:21:15.679 Casie Aviles: Probably wrong.

251 00:21:16.090 00:21:18.040 Samuel Roberts: Yeah, but if it’s knowledge.

252 00:21:18.040 00:21:25.400 Pranav Narahari: But that is fair. They do use, like, terms like knowledge base, and then… not necessarily, like, they just might want to create, like, an additional…

253 00:21:25.620 00:21:31.820 Pranav Narahari: way to, like, add an additional file as context, and if that’s the case… Very different, right?

254 00:21:31.820 00:21:36.729 Samuel Roberts: Exactly, that’s kind of why, like, it’s a wide range of anywhere from, like, you know.

255 00:21:36.870 00:21:40.580 Samuel Roberts: Trivially throwing in some more context and adding a table.

256 00:21:40.700 00:21:41.719 Samuel Roberts: entry that has.

257 00:21:42.600 00:21:44.760 Samuel Roberts: Some text that we can inject into the prompt.

258 00:21:44.950 00:21:50.669 Samuel Roberts: To, yeah, a full embedding table where everything gets stored, and…

259 00:21:51.220 00:21:53.969 Samuel Roberts: Yeah. Per brand, and injected, and searched.

260 00:21:54.260 00:22:00.420 Pranav Narahari: I totally agree. I think… What we can do here is, like, In the most, like.

261 00:22:00.920 00:22:04.699 Pranav Narahari: Because my assumption is just that that’s not the case, that they’re going to.

262 00:22:04.700 00:22:05.179 Samuel Roberts: Yeah, I would do.

263 00:22:05.180 00:22:08.960 Pranav Narahari: bunch of data here, and it’s, like, it’s not feasible, actually, to, like.

264 00:22:09.520 00:22:14.349 Pranav Narahari: add that… all of that in just the raw format as context. So…

265 00:22:14.350 00:22:15.860 Samuel Roberts: If it’s a lot of stuff, yeah, totally.

266 00:22:15.860 00:22:22.019 Pranav Narahari: Yeah, there’s a ton of stuff, and I assume, like, for these brands, it will be.

267 00:22:23.600 00:22:27.660 Pranav Narahari: So, what I’ll do is, though, I will, like, like,

268 00:22:28.300 00:22:35.179 Pranav Narahari: I will check to see, like, what they… what exactly they need in our call, like, when I give up this estimate.

269 00:22:35.490 00:22:37.629 Pranav Narahari: I’m thinking, though, just based on…

270 00:22:38.190 00:22:40.730 Pranav Narahari: What we know here is that…

271 00:22:40.910 00:22:44.359 Pranav Narahari: This is gonna take, like, a lot of hours. Yeah.

272 00:22:44.690 00:22:45.830 Pranav Narahari: So…

273 00:22:47.780 00:22:56.679 Pranav Narahari: Okay. Yeah, because I’m thinking, like, this is not even on a one-week, maybe not even a two-week scale, like, we… to actually get this into production, like, it could take 3 weeks.

274 00:22:57.390 00:23:00.230 Pranav Narahari: And when I say 3 weeks, I mean 120 hours.

275 00:23:00.880 00:23:07.189 Samuel Roberts: Yeah, yeah, easily, easily. I think that’s… if that’s what they… if they want the whole, you know, more complex side of it, definitely.

276 00:23:07.270 00:23:14.799 Pranav Narahari: If they’re looking for just, like, the AI doesn’t know anything about the brand, and we want to give it a little more context, which is the other end of the spectrum.

277 00:23:14.800 00:23:19.320 Samuel Roberts: Like, we could easily get that in there, and they could test it out, and see how it is, you know?

278 00:23:19.320 00:23:19.790 Pranav Narahari: Yeah.

279 00:23:19.790 00:23:23.069 Samuel Roberts: That might do, like, 80% of what they need, kind of thing.

280 00:23:23.770 00:23:28.399 Pranav Narahari: Well, one other thing, too, is, like, if we… if they even need to save, like, certain…

281 00:23:29.400 00:23:36.989 Pranav Narahari: like, documents for a brand, like, we won’t be able to really save that in a Postgres table, right? Like, we’re gonna have to, like.

282 00:23:37.600 00:23:39.680 Pranav Narahari: What, are we gonna have to set up S3?

283 00:23:41.620 00:23:48.629 Samuel Roberts: Yeah, I mean, if they’re uploading PDFs and stuff, we definitely need to store that somewhere else.

284 00:23:48.630 00:23:49.540 Pranav Narahari: Yeah.

285 00:23:49.580 00:23:53.080 Samuel Roberts: We do have a bucket already set up.

286 00:23:53.850 00:23:58.450 Samuel Roberts: For the images, so it’s not… Crazy.

287 00:23:59.120 00:24:00.730 Samuel Roberts: But if they’re just… Oh, okay.

288 00:24:01.290 00:24:03.289 Samuel Roberts: If they’re just looking for, you know…

289 00:24:03.720 00:24:05.770 Samuel Roberts: Text, like, that could be stored.

290 00:24:08.160 00:24:09.860 Samuel Roberts: That could be stored in Superbase.

291 00:24:10.380 00:24:12.840 Samuel Roberts: If they’re doing… I mean, if we’re doing embeddings, yeah, we…

292 00:24:14.620 00:24:18.449 Samuel Roberts: We want to store the PDFs and embed them, so yeah, we would need both.

293 00:24:19.640 00:24:20.220 Pranav Narahari: Yep.

294 00:24:20.940 00:24:25.910 Pranav Narahari: Okay, cool. So then I’ll just leave this, you know, maybe I’ll just talk to them more to… before I create.

295 00:24:25.910 00:24:28.429 Samuel Roberts: I would… I would run them through the full, like.

296 00:24:28.980 00:24:39.260 Samuel Roberts: it’s, you know, full quote-unquote knowledge base with RAG, pipeline, and lookup, and multi-doc, and multimodal is… Yep.

297 00:24:39.730 00:24:45.649 Samuel Roberts: big, but again, I think it’s an 80-20 thing here, where I don’t know…

298 00:24:46.890 00:24:53.850 Samuel Roberts: What they… if they don’t realize how big that is, and how it’s probably overkill, depending on exactly what they need.

299 00:24:54.100 00:25:00.859 Samuel Roberts: That, you know, a knowledge base per brand eventually makes sense, but if it is just, like…

300 00:25:01.410 00:25:04.069 Samuel Roberts: A little bit of context per brand.

301 00:25:04.330 00:25:06.799 Samuel Roberts: And that gets them a chunk of it.

302 00:25:07.190 00:25:08.410 Samuel Roberts: done now.

303 00:25:08.650 00:25:14.099 Samuel Roberts: You know, maybe you can convince them that that’s worth it, and easier, and shorter, and cheaper.

304 00:25:15.940 00:25:16.500 Pranav Narahari: Understood.

305 00:25:16.500 00:25:17.340 Samuel Roberts: they need.

306 00:25:17.900 00:25:22.850 Pranav Narahari: Yeah. I mean, part of it for me is, like, I’d rather take on more work.

307 00:25:24.260 00:25:31.589 Pranav Narahari: I don’t know how you think about it, like, I’d like to, like… I mean, we have to accurately, like, assess how much time it would take, but…

308 00:25:32.810 00:25:44.630 Samuel Roberts: Well, what I’m saying is, like, I think to start, like, we could get, like, instead of saying, like, we’re gonna build this knowledge base, and it’s gonna take 3 weeks, and you’re not gonna see any, you know what I mean? Like, the iterative cycle is much longer then.

309 00:25:44.630 00:25:45.000 Pranav Narahari: Yeah.

310 00:25:45.000 00:25:53.170 Samuel Roberts: Whereas, if we could be like, oh, here’s some context, you guys could be using it while we’re finishing up any of these other pieces.

311 00:25:53.740 00:26:02.619 Samuel Roberts: It makes more sense to say, like, okay, well, then we can also spec out what a full knowledge base is, we can get a much better estimate, we can understand the actual, like.

312 00:26:02.760 00:26:04.150 Samuel Roberts: Requirements.

313 00:26:04.560 00:26:04.920 Pranav Narahari: Yeah.

314 00:26:04.920 00:26:05.500 Samuel Roberts: of…

315 00:26:05.610 00:26:09.709 Samuel Roberts: Like, what they’re trying to get out of it, rather than just, like, we want a knowledge base, you know?

316 00:26:11.000 00:26:19.349 Samuel Roberts: I think, like, it just gives them, like, an ability to have something to work with, if that’s what they need, while we build out a bigger piece.

317 00:26:20.530 00:26:21.350 Pranav Narahari: Yeah.

318 00:26:22.310 00:26:28.410 Samuel Roberts: Yeah, I get what you’re saying, like, I’m not trying to be like, it’s less work, we should do it this way, I’m just… I don’t know if they need…

319 00:26:29.090 00:26:35.260 Samuel Roberts: everything up, like… You know, if we could do something like that in a couple days, Versus…

320 00:26:35.550 00:26:42.160 Samuel Roberts: not do it, and then have the 3-week wait period, at least, you know what I mean? That’s 3 weeks of work, not just, you know, calendar time.

321 00:26:42.510 00:26:53.380 Pranav Narahari: Gotcha. Okay, one question I have, then, is, let’s say it’s… I kind of… we’ve kind of defined, like, the most complex to the least complex. Even for the least complex, I think this takes a week.

322 00:26:53.980 00:26:59.240 Pranav Narahari: You guys have a little bit more experience, I think, with just, like.

323 00:26:59.360 00:27:09.410 Pranav Narahari: setting up that, S3 bucket, I think you guys did it for, or Casey, you might have done it for the image generation tool, right?

324 00:27:10.240 00:27:14.319 Pranav Narahari: Do you think that estimate sounds about right? Like, 40 hours?

325 00:27:14.550 00:27:17.000 Casie Aviles: Yeah, I think Alicia’s… It’s good.

326 00:27:17.540 00:27:18.150 Pranav Narahari: Okay.

327 00:27:18.510 00:27:19.150 Samuel Roberts: Yeah.

328 00:27:19.790 00:27:20.540 Pranav Narahari: Okay, cool.

329 00:27:22.950 00:27:36.280 Pranav Narahari: Alright, well, yeah, I think… I feel pretty good about just, like, letting them know. So let’s maybe just do, like, an overall number as well, just so, like, we’re on the same page there. So we said, yeah, 40, 40,

330 00:27:37.040 00:27:43.160 Pranav Narahari: Meta dashboard review… I think we said

331 00:27:43.960 00:27:48.029 Pranav Narahari: Yeah, that would take just 16 hours,

332 00:27:49.180 00:27:51.889 Pranav Narahari: Prompt management… what did we say?

333 00:27:52.090 00:27:55.400 Pranav Narahari: Sorry, my brain is… Not working right now.

334 00:27:57.970 00:27:59.460 Casie Aviles: Meta dashboard.

335 00:27:59.830 00:28:03.529 Casie Aviles: I think we… we set it to 16, right, or 24.

336 00:28:03.730 00:28:05.150 Pranav Narahari: Yeah, that one’s 16, I think we said.

337 00:28:05.150 00:28:09.230 Samuel Roberts: That was my… yeah. I would say prompt management…

338 00:28:09.450 00:28:12.019 Samuel Roberts: So we gotta have tagging, we gotta have the…

339 00:28:12.190 00:28:16.139 Samuel Roberts: The scheduling… the Slack is the only thing I’m not…

340 00:28:16.600 00:28:19.049 Samuel Roberts: confident I know how long that’ll take, so I would…

341 00:28:19.850 00:28:22.809 Samuel Roberts: edge higher than I’m probably thinking, even so.

342 00:28:26.050 00:28:26.900 Pranav Narahari: Gotcha.

343 00:28:26.900 00:28:33.560 Samuel Roberts: Like, I think we could do, like, the global… like, this is the other piece where, like, yeah, total prompt management might be, like, I don’t know, 60…

344 00:28:34.090 00:28:34.710 Pranav Narahari: Yeah.

345 00:28:34.950 00:28:37.610 Pranav Narahari: But… I was gonna say 62, actually, so…

346 00:28:37.610 00:28:43.340 Samuel Roberts: So, if we split that out, like, we could easily add the tags. We could, you know, we could phase it.

347 00:28:43.760 00:28:44.300 Pranav Narahari: Oh, totally.

348 00:28:44.300 00:28:49.180 Samuel Roberts: So that’s why I just wanted that, as long as it’s like, yeah, like, it’s not gonna be 60 hours for them to get…

349 00:28:49.890 00:28:52.330 Samuel Roberts: The global versus specific brands kind of thing.

350 00:28:52.330 00:28:52.700 Pranav Narahari: Yeah.

351 00:28:52.700 00:28:57.530 Samuel Roberts: Look at the scheduling part of it, like, you know, you need that first piece to get the other piece kind of thing, so…

352 00:28:57.530 00:29:12.260 Pranav Narahari: What they’re gonna look at is they’re gonna just look at the total amount, and then they’ll just see what’s shipped after, like, by that certain date. Yeah. Yeah. But then, also, while we’re working on this, they’re gonna expect things to be demoed weekly, and so that’ll still be there.

353 00:29:12.690 00:29:15.320 Samuel Roberts: Yeah, I think that’s where these gotta get broken down, like, these are massive.

354 00:29:15.320 00:29:18.079 Pranav Narahari: 100%, and yeah, we’ll definitely do that.

355 00:29:18.270 00:29:18.940 Samuel Roberts: Yeah, that amount.

356 00:29:20.320 00:29:28.069 Pranav Narahari: Okay, and then as knowledge bases per brand, that is 40+. That could be way more than 40. At the minimum, it’s gonna be 40.

357 00:29:28.220 00:29:29.100 Samuel Roberts: Yeah…

358 00:29:29.570 00:29:32.950 Pranav Narahari: Okay, cool. So then that is… Yeah.

359 00:29:33.270 00:29:40.840 Pranav Narahari: That’s basically, okay, 4… 120 plus… 60… It’s basically $200.

360 00:29:41.130 00:29:41.920 Pranav Narahari: Okay.

361 00:29:44.170 00:29:48.050 Pranav Narahari: And then… Yeah, so kind of to give you guys…

362 00:29:48.170 00:29:55.429 Pranav Narahari: I had an idea of, like, okay, 200 hours, how long that would take for, like, our, like, optimum velocity would be…

363 00:29:56.530 00:29:59.510 Pranav Narahari: We would want to be spending around…

364 00:29:59.960 00:30:04.819 Pranav Narahari: 100 hours per month? 100 to 125 hours per month.

365 00:30:05.220 00:30:06.890 Pranav Narahari: Like, across the three of us.

366 00:30:07.680 00:30:08.240 Samuel Roberts: Okay.

367 00:30:08.360 00:30:09.020 Pranav Narahari: Yeah.

368 00:30:09.660 00:30:13.060 Pranav Narahari: And then, so, that would basically be, like, one and a half to two months.

369 00:30:15.310 00:30:16.469 Pranav Narahari: I think that’s fair.

370 00:30:16.760 00:30:17.330 Samuel Roberts: Okay.

371 00:30:18.350 00:30:22.500 Samuel Roberts: So, 100 hours… per month.

372 00:30:23.110 00:30:24.509 Samuel Roberts: Across the three of us?

373 00:30:25.160 00:30:25.830 Pranav Narahari: Yeah.

374 00:30:26.280 00:30:27.969 Pranav Narahari: 100 to 125.

375 00:30:28.170 00:30:32.280 Samuel Roberts: Oh, I see what you’re saying. Yeah, okay, I got enough. I was just trying to… I was catching up in the math in my head. Okay.

376 00:30:34.850 00:30:36.420 Samuel Roberts: And this is total how much?

377 00:30:36.600 00:30:37.650 Samuel Roberts: Where’d it come to?

378 00:30:38.990 00:30:41.599 Pranav Narahari: This at $200, a little bit less. Okay. Like, 190.

379 00:30:41.600 00:30:42.630 Samuel Roberts: Alright, that sounds right.

380 00:30:43.060 00:30:43.580 Pranav Narahari: Yeah.

381 00:30:43.870 00:30:45.260 Pranav Narahari: Okay, cool.

382 00:30:46.580 00:30:49.010 Pranav Narahari: Alright, thank you guys.

383 00:30:49.010 00:30:49.510 Samuel Roberts: drain.

384 00:30:49.510 00:31:00.949 Pranav Narahari: I’ll let you guys know what Bobby says, and then… yeah, I totally agree, Sam, like, next step is then breaking this down into, like, bite-sized tasks, creating the Gantt chart. I just want to first see, like, what they sign off on.

385 00:31:01.220 00:31:05.239 Samuel Roberts: Yeah, especially, like, the bigger ones, like the Knowledge Base and the,

386 00:31:06.850 00:31:11.449 Samuel Roberts: Blanking now, hold on, let me jump back over. The prompt management, like, breaking that into pieces.

387 00:31:11.950 00:31:13.190 Pranav Narahari: Totally. Yup.

388 00:31:14.410 00:31:14.960 Samuel Roberts: Cool.

389 00:31:15.320 00:31:19.709 Pranav Narahari: Alright guys, I’m gonna jump, but yeah, thank you guys for all the help, I’ll keep you guys updated.

390 00:31:20.100 00:31:21.389 Samuel Roberts: Alright, yeah, sounds good.

391 00:31:21.830 00:31:22.850 Pranav Narahari: Yep, see you guys.