Meeting Title: Lilo Team Sync Date: 2026-02-02 Meeting participants: Casie Aviles, Samuel Roberts, Pranav Narahari
WEBVTT
1 00:01:17.950 ⇒ 00:01:18.960 Casie Aviles: Hey, sound.
2 00:01:19.720 ⇒ 00:01:21.350 Samuel Roberts: Hey, sorry.
3 00:01:23.480 ⇒ 00:01:30.759 Samuel Roberts: Ugh, man, alright. Lot of… A lot of context switching already today.
4 00:01:31.220 ⇒ 00:01:34.970 Casie Aviles: Yeah, I know. I also have a bunch of meetings lined up.
5 00:01:35.500 ⇒ 00:01:38.289 Samuel Roberts: Yeah, Mondays… Mondays is kind of like that, I guess, so…
6 00:01:41.140 ⇒ 00:01:41.510 Pranav Narahari: Hey, guys.
7 00:01:41.510 ⇒ 00:01:42.400 Casie Aviles: Okay.
8 00:01:42.400 ⇒ 00:01:43.290 Samuel Roberts: Age.
9 00:01:46.010 ⇒ 00:01:48.149 Pranav Narahari: Are you guys getting a lot of buyers from land?
10 00:01:50.080 ⇒ 00:01:53.240 Samuel Roberts: No, I don’t… I don’t hear anything.
11 00:01:53.410 ⇒ 00:01:59.919 Pranav Narahari: You don’t? Okay, good. Because I just have the washing machine going, and it’s pretty loud. Okay. Yeah, no, the… I found that…
12 00:02:00.050 ⇒ 00:02:05.699 Samuel Roberts: I was actually just on FaceTime with my sister, and, like, she could hear my keyboard, but, like, I don’t even know if it picks up my keyboard here.
13 00:02:06.390 ⇒ 00:02:08.080 Pranav Narahari: Oh yeah, I’ve never heard your two word.
14 00:02:12.270 ⇒ 00:02:13.609 Samuel Roberts: Can you hear me typing just now?
15 00:02:14.120 ⇒ 00:02:15.300 Pranav Narahari: No. Yeah.
16 00:02:15.300 ⇒ 00:02:18.680 Samuel Roberts: Yeah, Zoom’s… Zoom’s noise cancellation is… Incredible.
17 00:02:18.800 ⇒ 00:02:20.599 Samuel Roberts: Yeah, that’s impressive. Okay, cool.
18 00:02:21.060 ⇒ 00:02:24.099 Samuel Roberts: Because, like, FaceTime was like, she’s like, what are you doing? I’m like, I’m just typing.
19 00:02:28.160 ⇒ 00:02:28.900 Samuel Roberts: Alright.
20 00:02:30.470 ⇒ 00:02:32.260 Casie Aviles: Okay,
21 00:02:33.130 ⇒ 00:02:42.909 Casie Aviles: Yeah, so I just have this, I think my… I guess my agenda for today is to just make sure that the Gantt and the tickets are up to date, and
22 00:02:43.080 ⇒ 00:02:52.510 Casie Aviles: If we think we can… Get our, you know, the tasks that we have for this week done.
23 00:02:54.340 ⇒ 00:03:00.249 Casie Aviles: Okay, cool, so I’ll just start with the Gantt chart. So, I believe for this… these two…
24 00:03:00.630 ⇒ 00:03:05.069 Casie Aviles: tasks here. We should be good here, right, I believe.
25 00:03:07.850 ⇒ 00:03:10.860 Casie Aviles: Or are we missing anything in particular?
26 00:03:12.130 ⇒ 00:03:16.499 Samuel Roberts: I mean, the warehouse is… I mean, I just feel like we… we kind of got the…
27 00:03:17.590 ⇒ 00:03:25.749 Samuel Roberts: warehouse started? I don’t know if I… like, do we… do we give any more context for warehouse setup at all? Like, what we were considering done for that?
28 00:03:27.390 ⇒ 00:03:27.870 Casie Aviles: I see.
29 00:03:27.870 ⇒ 00:03:35.370 Samuel Roberts: Yeah, okay. Like, I’m just thinking, like, we got Shopify syncing for one brand into the data warehouse, so that probably qualifies that.
30 00:03:35.530 ⇒ 00:03:41.570 Samuel Roberts: That it meant we got mother ducks set up, that meant we got, a polyatomic…
31 00:03:42.630 ⇒ 00:03:45.459 Samuel Roberts: So, like, we’re… we’re kind of ready there.
32 00:03:46.950 ⇒ 00:03:50.610 Samuel Roberts: I just think there’s some outstanding stuff. So I, I mean, I guess it’s, yeah, phase one…
33 00:03:51.650 ⇒ 00:03:58.669 Samuel Roberts: Yep, that’s fine. Data warehouse decision, that’s good. Phase 1 completion, really deployment and migration, auth setup.
34 00:04:01.330 ⇒ 00:04:04.460 Pranav Narahari: Yeah, that Phase 1… I don’t know if that’s really relevant, to be honest.
35 00:04:04.460 ⇒ 00:04:09.519 Samuel Roberts: Yeah, I’m just thinking, I feel like we… we’ve got it… Set up in that sense.
36 00:04:09.690 ⇒ 00:04:13.880 Pranav Narahari: Yeah, because that was, like, if we’re gonna do a self-hosted solution, we’re using cloud, so…
37 00:04:14.230 ⇒ 00:04:16.410 Samuel Roberts: Sure, yeah, no, no, no, I think that’s fine.
38 00:04:17.350 ⇒ 00:04:21.239 Samuel Roberts: The only thing that’s, like, outstanding in my mind about the warehouse…
39 00:04:21.740 ⇒ 00:04:26.440 Samuel Roberts: and this is not necessarily something captured by this ticket, I just want to kind of say it out loud, is the, like.
40 00:04:26.850 ⇒ 00:04:28.590 Samuel Roberts: How we’re gonna scale up.
41 00:04:28.980 ⇒ 00:04:33.140 Samuel Roberts: All the brands, and all the… connectors, right?
42 00:04:33.890 ⇒ 00:04:34.490 Pranav Narahari: Yeah.
43 00:04:35.360 ⇒ 00:04:39.999 Samuel Roberts: I looked a little bit at Polytomic’s docs, and I put together a little bit of a plan, and
44 00:04:42.480 ⇒ 00:04:43.650 Samuel Roberts: and cursor.
45 00:04:44.260 ⇒ 00:04:46.039 Samuel Roberts: But I think we…
46 00:04:46.870 ⇒ 00:04:53.540 Samuel Roberts: Need to think that through a little bit more, and, like, how we’re gonna manage all that once it’s up and running, too.
47 00:04:53.990 ⇒ 00:05:02.789 Pranav Narahari: Yeah, like, part of this POC could also be just, like, you know, we’re not gonna try doing it with all 70 brands right now, but maybe we try doing it with 2, just so we can.
48 00:05:02.790 ⇒ 00:05:03.619 Samuel Roberts: And she, like…
49 00:05:03.620 ⇒ 00:05:10.949 Pranav Narahari: what does that look like in Mother Duck? Like, how are we then gonna pull in the dashboard data dynamically?
50 00:05:11.970 ⇒ 00:05:16.759 Samuel Roberts: Right, so there’s definitely that, yeah, there’s that, and then there’s just, like, every time a brand gets added.
51 00:05:17.360 ⇒ 00:05:20.199 Samuel Roberts: what happens to set up a Shopify sync, you know?
52 00:05:20.390 ⇒ 00:05:21.870 Pranav Narahari: Oh, I see, yeah, that, that’s…
53 00:05:21.870 ⇒ 00:05:30.459 Samuel Roberts: what happens to the auth, what happens to the… like, this… some of that stuff, I think we probably need to add to the… either the Gantt or linear, just to make sure that, like, a spike happens on that.
54 00:05:32.790 ⇒ 00:05:37.489 Samuel Roberts: Otherwise, like, I would consider this done, yeah. I think we’re done in terms of, like, we got Mother Duck, we got…
55 00:05:37.710 ⇒ 00:05:39.720 Pranav Narahari: I mean, we’ve got Polytomic and AirByte now.
56 00:05:39.840 ⇒ 00:05:41.080 Samuel Roberts: Whichever we need.
57 00:05:43.590 ⇒ 00:05:44.330 Casie Aviles: Okay.
58 00:05:45.030 ⇒ 00:05:48.379 Samuel Roberts: So I was… yeah, I think we’re good there. I think that’s fine. I just wanted to…
59 00:05:48.530 ⇒ 00:05:52.170 Samuel Roberts: make this known, because I think, like, There’s definitely more to do.
60 00:05:53.940 ⇒ 00:05:56.729 Casie Aviles: I see, yeah, so… hmm, okay, okay.
61 00:05:57.320 ⇒ 00:06:02.089 Samuel Roberts: And then the daily pacing dashboard and the… Yeah, for PSC purposes, totally, totally, totally.
62 00:06:02.780 ⇒ 00:06:03.390 Casie Aviles: Alright.
63 00:06:03.390 ⇒ 00:06:08.349 Samuel Roberts: Like, formal data ingestion Pipelines might be kind of what I’m talking about here. Like, that might capture it well enough.
64 00:06:08.730 ⇒ 00:06:09.670 Samuel Roberts: You know? Okay.
65 00:06:10.330 ⇒ 00:06:13.149 Samuel Roberts: I just want to make sure that we said some of it out loud.
66 00:06:14.490 ⇒ 00:06:20.579 Samuel Roberts: And then, for daily pacing dashboard versus basic forecast and pacing model.
67 00:06:21.380 ⇒ 00:06:24.330 Samuel Roberts: How are we thinking of those as, like, didn’t we…
68 00:06:24.330 ⇒ 00:06:24.840 Pranav Narahari: Yeah.
69 00:06:24.840 ⇒ 00:06:26.979 Samuel Roberts: Talk about that a little bit in linear, I think? Okay.
70 00:06:26.980 ⇒ 00:06:32.859 Pranav Narahari: Yeah, yeah, so… these were kind of named before, like, there was a little bit of a scope change, so…
71 00:06:32.860 ⇒ 00:06:33.580 Samuel Roberts: Yeah, yeah.
72 00:06:33.670 ⇒ 00:06:43.009 Pranav Narahari: We talked a little bit about having, like, that, Orca, like, Colab notebook, and so that’s what I was kind of thinking of, like, okay, we need to build in, like, some script to…
73 00:06:43.270 ⇒ 00:06:47.059 Pranav Narahari: really just, like, differentiate all the data. I think it’s a lot.
74 00:06:47.060 ⇒ 00:06:47.380 Samuel Roberts: Absolutely.
75 00:06:47.380 ⇒ 00:06:57.150 Pranav Narahari: than that. It’s what I’m kind of saying is, like, the acceptance criteria for basic forecast and pacing model, which is… it’s probably just not, like, a super accurate title. It’s just, like.
76 00:06:57.300 ⇒ 00:07:04.420 Pranav Narahari: coming up with that SQL query to, like, track the relevant data from the… from Mother Duck.
77 00:07:04.600 ⇒ 00:07:07.940 Pranav Narahari: So that we can paste it into the UI, basically.
78 00:07:08.640 ⇒ 00:07:13.759 Samuel Roberts: Okay. Yeah. Okay, so I think that’s fine. I… the daily pacing dashboard, then, is probably…
79 00:07:13.760 ⇒ 00:07:16.089 Pranav Narahari: That’s just the UI, basically, and then being.
80 00:07:16.090 ⇒ 00:07:19.629 Samuel Roberts: Right, okay. Being connected to, like, Mother Duck, so it’s not…
81 00:07:19.830 ⇒ 00:07:22.899 Pranav Narahari: there yet, because the U… because the query, like.
82 00:07:23.190 ⇒ 00:07:29.409 Pranav Narahari: what Bobby said, like, in our meeting on Friday was, like, the data looks still a little bit off, so…
83 00:07:29.410 ⇒ 00:07:30.000 Samuel Roberts: I see.
84 00:07:30.000 ⇒ 00:07:31.320 Pranav Narahari: We were kind of coupled.
85 00:07:32.570 ⇒ 00:07:33.370 Samuel Roberts: Okay.
86 00:07:33.490 ⇒ 00:07:33.990 Pranav Narahari: Yeah.
87 00:07:36.040 ⇒ 00:07:38.210 Samuel Roberts: The data being off…
88 00:07:40.080 ⇒ 00:07:45.190 Samuel Roberts: How has that… I mean, I don’t want to derail this too much, but I… we can dig into that later, actually. Let’s keep going.
89 00:07:47.140 ⇒ 00:07:51.789 Casie Aviles: Yeah, okay, so… I guess we, as part of that.
90 00:07:52.200 ⇒ 00:07:55.549 Casie Aviles: Since we’re… we’re also going to try to match with…
91 00:07:55.690 ⇒ 00:08:00.040 Casie Aviles: I think that was from Shopify QL, right? What we’re… what we’re trying to match.
92 00:08:00.220 ⇒ 00:08:00.930 Casie Aviles: Right.
93 00:08:00.930 ⇒ 00:08:06.939 Pranav Narahari: Yeah. Well, what we’re trying to match isn’t actually Shopify QL, because that’s not what they’re using in Watergraph.
94 00:08:06.940 ⇒ 00:08:09.129 Samuel Roberts: Oh, it’s that other water graph, right.
95 00:08:09.310 ⇒ 00:08:09.770 Pranav Narahari: Yeah.
96 00:08:09.770 ⇒ 00:08:14.639 Casie Aviles: Like, Watergraph, apparently they’re using just the… just a regular, I think.
97 00:08:14.640 ⇒ 00:08:25.129 Pranav Narahari: requests API, so… Yeah, I just need to figure out, because there’s this other table called Transactions, which…
98 00:08:25.180 ⇒ 00:08:25.900 Samuel Roberts: Hmm, okay.
99 00:08:25.900 ⇒ 00:08:30.939 Pranav Narahari: Has other, maybe more… I think it’s just, like,
100 00:08:31.360 ⇒ 00:08:35.070 Pranav Narahari: They’re using that as a source of truth versus, orders.
101 00:08:35.210 ⇒ 00:08:40.379 Pranav Narahari: for whatever reason, that’s, like, their decision. Maybe it is more accurate, I don’t know. Interesting.
102 00:08:40.750 ⇒ 00:08:43.889 Pranav Narahari: And so, yeah, that’s basically… I’m gonna try to…
103 00:08:44.600 ⇒ 00:08:50.990 Pranav Narahari: build out the same query that I have right now, except using the data from the transactions table, and then I think that’ll do the trick.
104 00:08:52.450 ⇒ 00:08:53.220 Pranav Narahari: Yeah.
105 00:08:54.050 ⇒ 00:08:56.839 Samuel Roberts: Yeah, alright, that sounds good. Keep us updated on that.
106 00:08:58.110 ⇒ 00:09:09.569 Casie Aviles: Alright, I’m also wondering, like, where… I’m not sure if I have, like, access to that yet, or if I found… if I have the link anywhere. I also just want to take a quick look.
107 00:09:09.970 ⇒ 00:09:12.240 Casie Aviles: Like, I mean… Yeah, yeah.
108 00:09:13.290 ⇒ 00:09:14.720 Casie Aviles: I don’t remember.
109 00:09:14.720 ⇒ 00:09:22.280 Pranav Narahari: paste it in the group chat? If not, I can… I have it, for sure. I have it right here, let me just send it.
110 00:09:22.670 ⇒ 00:09:23.330 Casie Aviles: Okay.
111 00:09:24.240 ⇒ 00:09:25.400 Casie Aviles: Put it in the sent to you.
112 00:09:26.380 ⇒ 00:09:27.000 Casie Aviles: Cool.
113 00:09:29.730 ⇒ 00:09:30.679 Casie Aviles: Alright, thank you.
114 00:09:30.810 ⇒ 00:09:32.190 Pranav Narahari: Yep, just kidding.
115 00:09:37.310 ⇒ 00:09:43.809 Samuel Roberts: Right, yeah, yeah, thank you, yes. That’s the other thing we have, linear group doing on top of it. Appreciate that.
116 00:09:44.500 ⇒ 00:09:47.829 Casie Aviles: Right, so we’ll push this back a bit then, right?
117 00:09:50.020 ⇒ 00:09:53.329 Casie Aviles: Since we’re still going to… we’re still finalizing, like.
118 00:09:54.070 ⇒ 00:09:58.010 Casie Aviles: the data, and then we haven’t really connected it to the UI, right?
119 00:09:58.760 ⇒ 00:10:03.369 Samuel Roberts: Yeah, that’s probably… either that or make a new one that’s, like, tied them together, you know?
120 00:10:04.140 ⇒ 00:10:05.390 Casie Aviles: Oh, okay.
121 00:10:05.690 ⇒ 00:10:11.110 Samuel Roberts: I don’t… yeah, I mean, I don’t want to dictate how best to do it, but it might just be worth adding a new one.
122 00:10:12.520 ⇒ 00:10:14.790 Samuel Roberts: This is something I think we gotta get a little more…
123 00:10:17.290 ⇒ 00:10:21.410 Samuel Roberts: I don’t know, consistent about, is, like, linear versus instant.
124 00:10:22.130 ⇒ 00:10:22.810 Samuel Roberts: I feel like…
125 00:10:22.810 ⇒ 00:10:23.830 Casie Aviles: Hmm.
126 00:10:23.850 ⇒ 00:10:28.989 Samuel Roberts: They don’t… they shouldn’t mirror each other perfectly, because linear should be more granular, you know?
127 00:10:30.360 ⇒ 00:10:31.450 Pranav Narahari: Yeah, I agree.
128 00:10:31.450 ⇒ 00:10:38.099 Samuel Roberts: So, like, daily pacing dashboard might… we might want to continue, like, draw that out, but we should know that, like, UI is done…
129 00:10:38.420 ⇒ 00:10:42.759 Samuel Roberts: Like, dyeing them together as the next task is part of that thing.
130 00:10:43.000 ⇒ 00:10:43.980 Samuel Roberts: You know what I mean?
131 00:10:44.370 ⇒ 00:10:46.109 Pranav Narahari: Yeah, I totally agree.
132 00:10:46.550 ⇒ 00:10:53.130 Samuel Roberts: But I think that’s just… that’s something we’ll get better at over time. I don’t want to, like… I just figured I’d mention it now, since we’re talking about this, but, like…
133 00:10:54.130 ⇒ 00:10:56.989 Samuel Roberts: Like, this might be multiple tickets, is my thought, you know?
134 00:10:57.000 ⇒ 00:10:57.800 Pranav Narahari: But, yeah.
135 00:10:57.800 ⇒ 00:10:58.210 Casie Aviles: Yeah, yeah.
136 00:10:58.210 ⇒ 00:10:58.809 Samuel Roberts: That’s more probable.
137 00:10:58.810 ⇒ 00:10:59.320 Casie Aviles: that sneaky.
138 00:10:59.320 ⇒ 00:11:06.680 Samuel Roberts: Yeah, I think that’s… I think that’s fine, to drag that out there. POC complete, I think that’s good. ETL tool and connector setup is…
139 00:11:06.950 ⇒ 00:11:11.329 Samuel Roberts: Oh, I didn’t realize we had another one of those. Yeah, I think that’s… that’s maybe… that might be even more of the…
140 00:11:13.660 ⇒ 00:11:17.880 Samuel Roberts: the… How to do it programmatically, stuff.
141 00:11:17.880 ⇒ 00:11:18.480 Casie Aviles: Okay.
142 00:11:19.040 ⇒ 00:11:19.790 Pranav Narahari: Yeah.
143 00:11:21.650 ⇒ 00:11:23.579 Samuel Roberts: At least, like, Shopify, because…
144 00:11:23.740 ⇒ 00:11:25.250 Pranav Narahari: Yeah. If we can get…
145 00:11:25.250 ⇒ 00:11:28.139 Samuel Roberts: you know, understand the Shopify connector well.
146 00:11:28.550 ⇒ 00:11:32.699 Samuel Roberts: Understand it in a way that we can spin up a brand.
147 00:11:32.870 ⇒ 00:11:37.739 Samuel Roberts: keep that up to date. Maybe even add a dashboard that we can see in the platform or something?
148 00:11:39.620 ⇒ 00:11:46.570 Samuel Roberts: you know, I don’t want to, like, give… I don’t want to, like, overdo it into… Ditch, but…
149 00:11:46.680 ⇒ 00:11:48.560 Samuel Roberts: Something’s gotta manage that data, you know?
150 00:11:49.250 ⇒ 00:11:49.940 Samuel Roberts: Oh, okay.
151 00:11:50.460 ⇒ 00:11:53.570 Samuel Roberts: And, like, we probably want a top-level view of, like.
152 00:11:53.800 ⇒ 00:11:55.859 Samuel Roberts: Are the connectors working and stuff, so…
153 00:11:59.020 ⇒ 00:11:59.670 Pranav Narahari: And we also have the.
154 00:11:59.670 ⇒ 00:12:00.270 Samuel Roberts: Slightly.
155 00:12:00.430 ⇒ 00:12:01.770 Pranav Narahari: Oh, I don’t know if, yeah…
156 00:12:02.150 ⇒ 00:12:07.739 Pranav Narahari: Okay, so did you just create that, formal data ingestion pipelines one? Or was that there before?
157 00:12:08.110 ⇒ 00:12:09.010 Samuel Roberts: No, I was there.
158 00:12:09.260 ⇒ 00:12:10.570 Pranav Narahari: Oh, that was there, okay.
159 00:12:10.730 ⇒ 00:12:11.470 Samuel Roberts: Yeah.
160 00:12:11.470 ⇒ 00:12:12.730 Casie Aviles: Yeah, that was there.
161 00:12:12.970 ⇒ 00:12:14.080 Pranav Narahari: Oh, okay, cool.
162 00:12:14.630 ⇒ 00:12:21.499 Pranav Narahari: And yeah, subscription brand handling, that’s the one I was talking to Bobby about. That one’s actually not in
163 00:12:21.600 ⇒ 00:12:25.790 Pranav Narahari: for this, so… We can actually… Yeah, we’re just trash.
164 00:12:27.200 ⇒ 00:12:28.129 Samuel Roberts: Whatever it is.
165 00:12:33.190 ⇒ 00:12:41.120 Samuel Roberts: Yeah, cool. Yeah, I think he said that there’s, like, tags and stuff we can use in the Shopify without having to connect to Recharge, so there might be something we can at least, like.
166 00:12:41.930 ⇒ 00:12:42.660 Samuel Roberts: You know.
167 00:12:43.350 ⇒ 00:12:45.070 Samuel Roberts: V-Zero it, kind of thing.
168 00:12:47.370 ⇒ 00:12:47.940 Pranav Narahari: Cool.
169 00:12:49.300 ⇒ 00:12:50.270 Samuel Roberts: This looks pretty good.
170 00:12:53.510 ⇒ 00:12:54.480 Casie Aviles: Mmm.
171 00:12:54.950 ⇒ 00:13:03.039 Casie Aviles: Alright, what else do we need? So, we don’t really need the paid accounts yet, right? We don’t need it for this week, right?
172 00:13:04.480 ⇒ 00:13:07.840 Casie Aviles: Because I think it’s just the POC that we want to complete.
173 00:13:08.060 ⇒ 00:13:08.670 Casie Aviles: Or…
174 00:13:08.670 ⇒ 00:13:15.540 Samuel Roberts: I mean, we don’t need it. I think that’s where, if we’re gonna start trying to…
175 00:13:16.870 ⇒ 00:13:21.549 Samuel Roberts: Automate connections of connectors and things, we probably do… soon?
176 00:13:21.720 ⇒ 00:13:26.560 Samuel Roberts: Like, for… if… If this one here… Is gonna be, like…
177 00:13:27.180 ⇒ 00:13:30.780 Samuel Roberts: setting up more Shopify and figuring out
178 00:13:31.190 ⇒ 00:13:34.060 Samuel Roberts: How we can do that programmatically and keeping track of things.
179 00:13:34.690 ⇒ 00:13:37.700 Samuel Roberts: We probably want to be doing that on their polyatomic…
180 00:13:38.800 ⇒ 00:13:39.740 Samuel Roberts: At least.
181 00:13:41.010 ⇒ 00:13:46.689 Pranav Narahari: Yeah, are we paying for that right now? Because I think, Casey, you have it set up on, like, the Brainforge one, right?
182 00:13:46.690 ⇒ 00:13:54.159 Samuel Roberts: Yeah, that’s… that’s why… did you see Utam freak out when I was, like, I said something, and he was like, don’t do that on the… on ours, and I was like, oh, that’s not what I’m talking about. And I was like.
183 00:13:54.160 ⇒ 00:13:55.570 Casie Aviles: Disabled it now.
184 00:13:55.680 ⇒ 00:14:05.359 Casie Aviles: I disabled it. Okay. But I think what I can do is, I can ask, Ghalib from Polytomic to create, like, a free trial version for us.
185 00:14:06.650 ⇒ 00:14:11.940 Samuel Roberts: Okay, I would… Yeah, if that’s the best way to get them set up with an account.
186 00:14:12.900 ⇒ 00:14:15.200 Samuel Roberts: That they’ll eventually pay for.
187 00:14:15.430 ⇒ 00:14:17.659 Samuel Roberts: I just want to make sure they’re aware of, like.
188 00:14:18.260 ⇒ 00:14:21.960 Samuel Roberts: the added costs of, eventually, Mother Duck and Polytomic.
189 00:14:23.620 ⇒ 00:14:27.009 Samuel Roberts: So that might be a thing to discuss today, CSO-wise.
190 00:14:27.870 ⇒ 00:14:29.890 Samuel Roberts: As well as the ad hoc stuff.
191 00:14:30.770 ⇒ 00:14:32.560 Pranav Narahari: Sounds good. I don’t know, cause I don’t know if that was…
192 00:14:32.560 ⇒ 00:14:35.099 Samuel Roberts: Like, laid out in the… in the,
193 00:14:35.310 ⇒ 00:14:39.840 Samuel Roberts: the PRD super well, because we didn’t know what tools we were going to be using yet, but just make sure that, like.
194 00:14:40.190 ⇒ 00:14:43.680 Samuel Roberts: However that’s sorted out, if it’s something we’re managing or they’re managing.
195 00:14:44.150 ⇒ 00:14:44.560 Pranav Narahari: Yeah.
196 00:14:44.560 ⇒ 00:14:45.359 Samuel Roberts: aware of it.
197 00:14:45.980 ⇒ 00:14:49.219 Pranav Narahari: I think what I’ll do is, I’m going to just get this, like.
198 00:14:49.680 ⇒ 00:14:52.859 Pranav Narahari: data thing crystal clear that, like.
199 00:14:53.070 ⇒ 00:15:00.729 Pranav Narahari: So then that’ll just basically solidify what tool exactly we’re using. Because all the way, everybody, they patched that.
200 00:15:01.070 ⇒ 00:15:05.150 Samuel Roberts: I saw that, yeah. I’m glad I commented, because I think that kick-started them a little bit.
201 00:15:05.870 ⇒ 00:15:08.410 Pranav Narahari: Yeah, yeah, yeah, that was, that was a good,
202 00:15:08.700 ⇒ 00:15:11.829 Pranav Narahari: It was good that we did that. It was cool that they kind of, you know…
203 00:15:12.360 ⇒ 00:15:15.539 Samuel Roberts: Yeah, it was just sitting there. I think it just hadn’t been, you know, tested and merged, so…
204 00:15:15.840 ⇒ 00:15:17.200 Samuel Roberts: Yeah, yeah.
205 00:15:18.040 ⇒ 00:15:22.550 Samuel Roberts: Yeah. But yeah, so we could trial that, but I mean… Yeah, either way.
206 00:15:23.990 ⇒ 00:15:28.389 Samuel Roberts: Yeah, you’re right. I just want to make sure that they’re aware that these things will have a cost, and we’ll have to figure out if we’re, like…
207 00:15:28.790 ⇒ 00:15:31.419 Samuel Roberts: Getting them to set up billing, or…
208 00:15:32.140 ⇒ 00:15:34.309 Samuel Roberts: However we’re doing it, but… Totally.
209 00:15:34.310 ⇒ 00:15:37.059 Pranav Narahari: Yeah, even if that is an agenda item on Friday.
210 00:15:37.160 ⇒ 00:15:47.150 Pranav Narahari: For, like, what we discuss, and then we can leave them that. Like, I… they don’t need to, like, do a ton of research, right? We’re just gonna… I’ll do the research on, like, Thursday.
211 00:15:47.150 ⇒ 00:15:48.150 Samuel Roberts: Or… Yeah.
212 00:15:48.740 ⇒ 00:15:50.680 Pranav Narahari: Yeah, let me…
213 00:15:51.480 ⇒ 00:15:57.950 Samuel Roberts: No, I’m just saying, like, in terms of, like, CSO, like, how do we normally communicate this to clients? What’s our normal path for…
214 00:15:58.340 ⇒ 00:16:07.070 Samuel Roberts: like, handling the ETL and the warehousing in terms of that. I just don’t… I haven’t really managed that yet on a client, so I wasn’t sure.
215 00:16:07.680 ⇒ 00:16:11.040 Pranav Narahari: Like… Gotcha. Yeah, that seems like CSO. I would…
216 00:16:11.040 ⇒ 00:16:17.429 Samuel Roberts: Yeah, and just, and just talk, yeah, with Utam, make sure, like, he’ll, he’ll know, because they’ve done this a bunch on other clients, so…
217 00:16:17.430 ⇒ 00:16:18.120 Pranav Narahari: Right.
218 00:16:19.320 ⇒ 00:16:20.300 Samuel Roberts: Sounds good.
219 00:16:22.650 ⇒ 00:16:23.869 Samuel Roberts: What else? Okay.
220 00:16:26.570 ⇒ 00:16:32.599 Samuel Roberts: Yeah, I mean, I think we’re looking pretty good out here for this. The ad hoc stuff’s the other thing, yeah, you gotta figure that out, too, with Lutama, about how we’re gonna handle…
221 00:16:32.870 ⇒ 00:16:34.250 Samuel Roberts: That.
222 00:16:34.250 ⇒ 00:16:37.730 Pranav Narahari: Yeah, I think for this time, like… we’ve, like…
223 00:16:39.390 ⇒ 00:16:42.300 Pranav Narahari: I’m not too concerned with,
224 00:16:43.170 ⇒ 00:16:51.370 Pranav Narahari: with Meta, specifically. And then Klaviyo, too, is like, this isn’t, this doesn’t have anything to do with the data warehouse.
225 00:16:51.370 ⇒ 00:16:51.779 Samuel Roberts: I think you’re right.
226 00:16:51.780 ⇒ 00:17:01.690 Pranav Narahari: And so, it’s really just kind of… it’s… I think it’s pretty lightweight, like this, ad hoc feature ad, and even the meta one is not super difficult, and from.
227 00:17:01.690 ⇒ 00:17:06.219 Samuel Roberts: I haven’t looked a ton at them, can you… can you, like, give me a TLDR for the meta dashboard?
228 00:17:06.220 ⇒ 00:17:18.779 Pranav Narahari: Yeah. Yeah. So, the meta dashboard is essentially just another, like, table, bringing in essentially the same data that we’re pulling into… for Slack reports.
229 00:17:19.220 ⇒ 00:17:28.650 Pranav Narahari: So, the Slack reports are already in production, they’re working, I have them scheduled, like, sent into, like, I see that it’s working every single day, so I’m not too worried about anything.
230 00:17:28.650 ⇒ 00:17:29.049 Samuel Roberts: So we can…
231 00:17:29.050 ⇒ 00:17:29.720 Pranav Narahari: That’s fair.
232 00:17:29.970 ⇒ 00:17:33.899 Samuel Roberts: So that can still be the API, and then we’ll flip that to the warehouse eventually, kind of thing?
233 00:17:34.540 ⇒ 00:17:41.900 Pranav Narahari: Yeah, like, I think… well, yeah, exactly, we can have it just as, like, the API for now, if needed.
234 00:17:42.090 ⇒ 00:17:59.439 Pranav Narahari: And it’s… it’s not gonna be, like, a heavy lift at all. What would be best case scenario is, like, yeah, like, if we’re ahead of schedule for this week, we can actually, like, do the whole, AirByte or Polytomic into Mother Duck for the… the Facebook ads library. Okay.
235 00:18:00.310 ⇒ 00:18:08.560 Pranav Narahari: Yeah, and then… all that ads data we can then use for the dashboard instead of the API.
236 00:18:10.740 ⇒ 00:18:14.840 Samuel Roberts: Okay, and the Klaviyo thing is that, like, scheduling thing?
237 00:18:15.320 ⇒ 00:18:19.390 Pranav Narahari: Basically, it’s, I think it’s kind of…
238 00:18:20.550 ⇒ 00:18:23.620 Pranav Narahari: Creating, like, email campaigns, and it’s, like, a tool.
239 00:18:24.150 ⇒ 00:18:26.230 Samuel Roberts: Right, so this is gonna be a tackle.
240 00:18:26.360 ⇒ 00:18:32.359 Pranav Narahari: you’ll have, like, a calendar of, like, when you want to, yeah, create. It’ll basically create, like.
241 00:18:32.490 ⇒ 00:18:39.019 Pranav Narahari: just 4 images, which are images of calendars, I think, of,
242 00:18:39.260 ⇒ 00:18:41.850 Pranav Narahari: Just, like, when you should, like, send out certain things.
243 00:18:42.910 ⇒ 00:18:47.950 Samuel Roberts: Is it… Are we gonna need a data model in our… Database for this?
244 00:18:52.320 ⇒ 00:18:53.329 Samuel Roberts: I just, I just, I haven’t looked…
245 00:18:53.750 ⇒ 00:18:54.929 Samuel Roberts: they did, so I’m not so sure.
246 00:18:54.930 ⇒ 00:18:57.650 Pranav Narahari: That’s a good question. I will look into it a little bit more, too.
247 00:18:58.180 ⇒ 00:19:05.630 Samuel Roberts: Okay. Yeah, I’m thinking… and we talked a little bit about, like, what is the standard procedure for these ad hoc things. I think there definitely should be a, like.
248 00:19:06.020 ⇒ 00:19:10.839 Samuel Roberts: Within whatever time period, we do a, like, spike on it.
249 00:19:13.260 ⇒ 00:19:14.690 Pranav Narahari: Yeah. And then…
250 00:19:14.690 ⇒ 00:19:20.050 Samuel Roberts: That way we’ll know, like, how much of a lift it is, what has to change, if it fits into other things.
251 00:19:20.760 ⇒ 00:19:21.310 Pranav Narahari: Totally.
252 00:19:22.420 ⇒ 00:19:35.979 Samuel Roberts: And then, like, that way we’re not committing to anything for them, but, like, we will definitely, within, like, you know, a week at the most, be like, here’s the plan for it. It’s either, yeah, we’ll fit it in now, we’ll fit it in later, it, you know, it’s gonna take these kinds of changes, like.
253 00:19:36.400 ⇒ 00:19:38.619 Samuel Roberts: Especially now, when there’s, like, not a ton…
254 00:19:39.740 ⇒ 00:19:43.429 Samuel Roberts: in terms of, like, Klaviyo stuff in the database yet, I just wasn’t sure if this is, like.
255 00:19:43.870 ⇒ 00:19:47.830 Samuel Roberts: Is Stitch managing the calendar? Is Klavo managing the calendar? We’re pulling from them?
256 00:19:48.580 ⇒ 00:19:49.390 Samuel Roberts: That’s all.
257 00:19:50.170 ⇒ 00:19:52.309 Pranav Narahari: Gotcha. Yeah. Let me see…
258 00:19:52.310 ⇒ 00:19:53.330 Samuel Roberts: I think that if we can…
259 00:19:53.870 ⇒ 00:20:03.990 Pranav Narahari: Yeah, okay. I didn’t put much description in these… I didn’t put any description in these, two tickets, so… let me… that’s something that I will sync up with them, like…
260 00:20:04.260 ⇒ 00:20:05.540 Pranav Narahari: Today on…
261 00:20:06.810 ⇒ 00:20:10.030 Samuel Roberts: Yeah, I would say just, like, a quick, yeah, a little spike of, like.
262 00:20:10.680 ⇒ 00:20:14.310 Samuel Roberts: What they did, what we need to do to make it work.
263 00:20:15.440 ⇒ 00:20:21.730 Samuel Roberts: And then the details from that, like, of if it’s API calls, what’s new, what’s already existing. Just, like, a quick, like…
264 00:20:22.040 ⇒ 00:20:26.889 Samuel Roberts: You know, and then we can estimate, like, oh yeah, this is a bigger lift than they think it is, or something.
265 00:20:27.140 ⇒ 00:20:27.800 Pranav Narahari: Right.
266 00:20:28.050 ⇒ 00:20:37.699 Samuel Roberts: Or not, you know, it might be even easier, who knows? But that’s… I just… without having had a chance to dig into those, I just… I think that’s probably a good path forward, especially if they’re going to be spitting out more of these things.
267 00:20:38.250 ⇒ 00:20:39.609 Pranav Narahari: 100%. Yeah.
268 00:20:42.440 ⇒ 00:20:43.060 Samuel Roberts: Okay.
269 00:20:45.960 ⇒ 00:20:47.790 Casie Aviles: Right. I think so.
270 00:20:47.790 ⇒ 00:20:52.750 Samuel Roberts: Yeah, and if you want me to take a look at anything, let me know, too. I just haven’t had a chance to look at those PRs or anything, so…
271 00:20:53.180 ⇒ 00:21:00.070 Pranav Narahari: Yeah, I think it’s fine. You can take a look at it, too, and just, I can get the alternate perspective,
272 00:21:00.670 ⇒ 00:21:07.039 Pranav Narahari: But… yeah, I remember them being pretty simple, which is why I was just like, yeah, we could definitely, like, bang these out.
273 00:21:07.040 ⇒ 00:21:07.640 Samuel Roberts: Okay.
274 00:21:08.000 ⇒ 00:21:13.900 Pranav Narahari: But… Because they also, like, created the UI, and it looked pretty good. Yeah.
275 00:21:14.130 ⇒ 00:21:15.250 Pranav Narahari: And…
276 00:21:15.920 ⇒ 00:21:29.279 Pranav Narahari: Yeah, so, the Klaviyo one is just, like, one that… I wasn’t fully understanding the use case of it. Is it, like, it seemed extremely simple, like, they just kind of wanted to create, like, calendar images based on, like.
277 00:21:29.440 ⇒ 00:21:32.660 Pranav Narahari: Like, a form that they fill out for, like, an email campaign.
278 00:21:34.160 ⇒ 00:21:39.919 Samuel Roberts: Oh, okay, because I know he posted some screencaps of something. Yeah, and I was just, like, I was asking specifically, like.
279 00:21:39.920 ⇒ 00:21:43.739 Pranav Narahari: what format do you want these images to look like? And that’s what he sent those images for.
280 00:21:45.260 ⇒ 00:21:46.779 Samuel Roberts: Are these images, or is this, like…
281 00:21:49.060 ⇒ 00:21:50.530 Pranav Narahari: Yeah, these are images.
282 00:21:51.070 ⇒ 00:21:51.520 Samuel Roberts: I mean.
283 00:21:52.600 ⇒ 00:21:54.750 Samuel Roberts: That’s what it’s gonna look like. I guess they’re also you.
284 00:22:01.580 ⇒ 00:22:04.470 Pranav Narahari: Yeah, okay, there’s definitely… I need to do some… I need to do a spike on this.
285 00:22:04.470 ⇒ 00:22:08.490 Samuel Roberts: Okay, yeah, do a little bit more digging, or yeah, do a spike, talk to them.
286 00:22:08.750 ⇒ 00:22:13.990 Samuel Roberts: We also might want to put in, like we were talking about even with, like, PRs for platform stuff, just like…
287 00:22:14.960 ⇒ 00:22:26.250 Samuel Roberts: have them do a loom walkthrough of what they did kind of thing, or, like, what they’re looking for as well. I don’t know what the PR descriptions are, maybe they’re fantastic, but a PR description and a loom maybe should be, like, a basic, like.
288 00:22:27.820 ⇒ 00:22:33.630 Samuel Roberts: Requirement for those, and then, like, that way we can probably do a spike without having to bother them too much.
289 00:22:34.080 ⇒ 00:22:35.180 Samuel Roberts: Most of the time.
290 00:22:38.140 ⇒ 00:22:40.719 Samuel Roberts: So, I’m just thinking process stuff right now, because my brain’s, like…
291 00:22:42.100 ⇒ 00:22:50.199 Samuel Roberts: Trying to figure out, like, all this platform stuff, too. I was talking to my sister, and she was like, can you get me set up on, like, an AI coding thing? And I was like, this is the third time today that someone’s been, like.
292 00:22:50.490 ⇒ 00:22:53.389 Samuel Roberts: We’re gonna have people that aren’t technical do coding, and I’m like, oh no.
293 00:22:54.800 ⇒ 00:22:55.849 Samuel Roberts: More of this.
294 00:22:56.140 ⇒ 00:23:03.689 Pranav Narahari: It’s definitely more complex and, like, could have a lot of downfalls than people think of, you know, like…
295 00:23:03.690 ⇒ 00:23:05.429 Samuel Roberts: That’s exactly… I was trying to…
296 00:23:05.670 ⇒ 00:23:19.269 Samuel Roberts: Yeah. Yeah. I was explaining to her that, like, it can get you 80% of the way there in what you think it’s doing, but then, like, that last 20% is all the, like, weird edge cases and, you know, bad paths that, like, you haven’t thought through.
297 00:23:19.800 ⇒ 00:23:24.669 Pranav Narahari: And my worry is just, like, if you move too fast and you just kind of create a bunch of slop, like…
298 00:23:25.030 ⇒ 00:23:28.219 Pranav Narahari: Then debugging that, like, can be, like, such a problem.
299 00:23:28.220 ⇒ 00:23:32.260 Samuel Roberts: That’s exactly what I said. I said I’m just gonna be reviewing AI code the rest of my life at this point.
300 00:23:32.520 ⇒ 00:23:33.120 Pranav Narahari: Yeah.
301 00:23:33.720 ⇒ 00:23:36.189 Samuel Roberts: Like, fixing up… yeah, it just was like…
302 00:23:36.370 ⇒ 00:23:38.910 Samuel Roberts: It was so funny that it was, like, the third time today someone was, like, talking.
303 00:23:39.490 ⇒ 00:23:46.259 Samuel Roberts: Like, oh boy. But I think, I think, yeah, as long as we get more processes in place for it, and, like, don’t just merge everything.
304 00:23:46.950 ⇒ 00:23:57.660 Samuel Roberts: Especially for Lilo specifically, like, they… not giving them access to the backend. At first, I was like, we should just let Claude go nuts and see what it does, but then I was like, you know what, maybe it’s better that they can’t.
305 00:24:00.250 ⇒ 00:24:03.399 Samuel Roberts: I just hope that handoff is gonna be clean.
306 00:24:03.550 ⇒ 00:24:04.060 Samuel Roberts: You know.
307 00:24:04.060 ⇒ 00:24:04.710 Pranav Narahari: Yeah.
308 00:24:05.030 ⇒ 00:24:08.690 Pranav Narahari: Actually, one quick question, completely unrelated.
309 00:24:08.690 ⇒ 00:24:09.270 Samuel Roberts: Yeah.
310 00:24:09.460 ⇒ 00:24:14.809 Pranav Narahari: Bobby sent, like, a message just, like, 30 minutes ago about, like, saving on API costs.
311 00:24:15.090 ⇒ 00:24:23.410 Pranav Narahari: Model selection will definitely help us a little bit there. Right now, I think we’re just using Sonnet port 4.5, right?
312 00:24:23.720 ⇒ 00:24:27.619 Samuel Roberts: Yeah, that’s it, I think I had started with Opus.
313 00:24:27.820 ⇒ 00:24:32.219 Pranav Narahari: And then I hard-coded that because that’s the only one that has a million…
314 00:24:32.970 ⇒ 00:24:33.960 Samuel Roberts: context.
315 00:24:34.350 ⇒ 00:24:35.590 Pranav Narahari: Context window? Okay.
316 00:24:35.880 ⇒ 00:24:39.619 Samuel Roberts: Gotcha. And I think it was… remember, they were hitting weird errors pretty quickly with them?
317 00:24:40.250 ⇒ 00:24:41.340 Samuel Roberts: Testing it?
318 00:24:41.700 ⇒ 00:24:50.319 Samuel Roberts: And they were like, last time we just flipped on this million dollar, or million token thing, and I was like, oh, okay, and I just dropped it to that. So, there’s definitely, like, where we could maybe use…
319 00:24:51.530 ⇒ 00:24:56.470 Samuel Roberts: Haiku for some things, or we could give them an option to flip, but then we’ll have to do some handling of, like.
320 00:24:56.760 ⇒ 00:25:00.199 Samuel Roberts: contact size. So we can, we can let them know that.
321 00:25:01.800 ⇒ 00:25:10.149 Samuel Roberts: Yeah, I’m just gonna message about that. Or go ahead. Yeah, that’s fine, yeah, just whatever I said. But there’s definitely a few things we could probably do with Haiku.
322 00:25:10.690 ⇒ 00:25:17.350 Samuel Roberts: like… like, the title generation right now, I think, is still Sonnets. Like, we could save a little bit there, you know what I mean? Like, there’s little places where, like, little things happen.
323 00:25:17.850 ⇒ 00:25:27.029 Pranav Narahari: Yeah, and I think we can also just, like, use Haiku when we can, and then if the context is too large, then we just switch to Sonnet.
324 00:25:27.720 ⇒ 00:25:29.209 Samuel Roberts: That might be a good idea, yeah.
325 00:25:29.210 ⇒ 00:25:33.769 Pranav Narahari: Yeah, I think that’s probably the best way to do it. I don’t really know how much it’s gonna save them.
326 00:25:34.510 ⇒ 00:25:42.239 Samuel Roberts: Yeah, I don’t actually know the pricing. I mean, I know Opus is, like… I was… just because I was playing with this weekend, I was like, oh, this is so much money, comparatively.
327 00:25:42.240 ⇒ 00:25:42.780 Pranav Narahari: Yeah.
328 00:25:42.920 ⇒ 00:25:45.470 Samuel Roberts: But, let me… Yeah.
329 00:25:46.180 ⇒ 00:25:55.999 Samuel Roberts: Yeah, I would just… I would just give them a little explainer, though, like what we just talked about, and just be like, yeah, we can bounce this around a little bit, the only issue is that context window that we found earlier.
330 00:25:57.210 ⇒ 00:26:00.360 Samuel Roberts: But yeah, we could also do some quick back-of-the-envelope calculations.
331 00:26:00.470 ⇒ 00:26:03.720 Samuel Roberts: Or… Mike.
332 00:26:03.950 ⇒ 00:26:05.980 Samuel Roberts: How much they could save, but…
333 00:26:09.460 ⇒ 00:26:11.480 Pranav Narahari: And the context window is mostly being…
334 00:26:11.670 ⇒ 00:26:13.930 Pranav Narahari: Sorry, I don’t want to completely derail this.
335 00:26:13.930 ⇒ 00:26:16.110 Samuel Roberts: But yeah, the content, though, is just, like.
336 00:26:16.250 ⇒ 00:26:22.619 Pranav Narahari: being used primarily or significantly by, like, the Claude Skills, or is that from…
337 00:26:22.810 ⇒ 00:26:28.460 Samuel Roberts: No, no, this was… I think it was the data coming back from the MCP servers.
338 00:26:28.460 ⇒ 00:26:29.070 Pranav Narahari: Yeah.
339 00:26:29.900 ⇒ 00:26:33.030 Samuel Roberts: It’s just huge JSON blobs.
340 00:26:33.280 ⇒ 00:26:33.850 Pranav Narahari: Yeah.
341 00:26:33.850 ⇒ 00:26:38.199 Samuel Roberts: that just… overwhelmed, what’s it?
342 00:26:38.820 ⇒ 00:26:42.140 Samuel Roberts: Sonic… Yeah.
343 00:26:44.960 ⇒ 00:26:54.429 Samuel Roberts: 1 million token context window, 5x increase from the 200,000 limit. So… Yeah, it’s just Sonnet.
344 00:26:55.130 ⇒ 00:26:59.940 Samuel Roberts: Yeah, for doing that kind of, like, MCP processing stuff, I think Sonnet’s the only way to go there.
345 00:27:00.720 ⇒ 00:27:03.610 Samuel Roberts: But… Like you said, like, we could…
346 00:27:04.810 ⇒ 00:27:11.240 Samuel Roberts: test that out a little bit more, make it so that Haiku is used by default, except when an MCP is called, or something else like that.
347 00:27:12.020 ⇒ 00:27:20.049 Pranav Narahari: Yeah, I think there’s probably a way to just do a quick calculation of how many tokens this is going to be, and then, based on that, we just route it to the right.
348 00:27:20.480 ⇒ 00:27:21.520 Pranav Narahari: Model.
349 00:27:21.980 ⇒ 00:27:25.249 Samuel Roberts: Yeah, I wonder, does the AI SDK have some good, like.
350 00:27:25.670 ⇒ 00:27:29.770 Pranav Narahari: Yeah, I think they have, like, an input tokens, like, field.
351 00:27:31.200 ⇒ 00:27:45.850 Samuel Roberts: Okay, yeah, I would say let them know that, and then we can do a little research into that, like, how we might implement that, if it’s necessary. Because I honestly, I have no idea how much they’re paying right now, you know, I have no insight into that, like, how many tokens we’ve run, and what they’re going to run on a normal basis.
352 00:27:49.890 ⇒ 00:27:50.800 Samuel Roberts: Oh, yeah.
353 00:27:51.590 ⇒ 00:27:58.580 Samuel Roberts: Okay, Haiku is, like, a million… a dollar per million, Sonnet 4-5 is $3 per million input.
354 00:28:00.740 ⇒ 00:28:02.809 Samuel Roberts: 15 per million output.
355 00:28:03.450 ⇒ 00:28:07.830 Samuel Roberts: Wow. Yeah, 5 per million output on Haiku. Okay, that’s a big difference.
356 00:28:12.510 ⇒ 00:28:19.719 Samuel Roberts: Okay, yeah, I would say, let them know we can look into that if that’s something that they, like, if that’s a, you know.
357 00:28:20.260 ⇒ 00:28:22.690 Samuel Roberts: I don’t know how to phrase it best, an issue, like a real issue.
358 00:28:24.160 ⇒ 00:28:29.060 Samuel Roberts: And then… but remind them about that context window. And again, we can probably even, like.
359 00:28:29.760 ⇒ 00:28:32.770 Samuel Roberts: Well, I don’t know, the MCPs go right to the model, don’t they?
360 00:28:33.430 ⇒ 00:28:34.909 Samuel Roberts: We could probably, like…
361 00:28:36.120 ⇒ 00:28:41.149 Samuel Roberts: dig into those MCPs and, like, strip down the data a little bit that comes back, maybe?
362 00:28:41.920 ⇒ 00:28:45.319 Samuel Roberts: I’ve seen some things about, like, other kinds of, like.
363 00:28:45.500 ⇒ 00:28:49.630 Samuel Roberts: structured data for LLMs, because, like, JSON can be very token heavy.
364 00:28:50.210 ⇒ 00:28:51.450 Pranav Narahari: Yeah, depending on.
365 00:28:51.450 ⇒ 00:28:53.820 Samuel Roberts: I just think of it as a good solution.
366 00:28:53.820 ⇒ 00:28:54.620 Pranav Narahari: Yeah.
367 00:28:54.770 ⇒ 00:28:55.570 Pranav Narahari: we…
368 00:28:55.570 ⇒ 00:29:03.719 Samuel Roberts: I’m saying, like, this is just things we can mention. I would just say, like, yeah, we can definitely look at that a little bit, just remember the token inputs, but if, like, cost is, you know.
369 00:29:05.700 ⇒ 00:29:11.049 Samuel Roberts: spiking or whatever, we can move to Haiku a bit more, except for calls with MCPs, maybe? I don’t know.
370 00:29:12.010 ⇒ 00:29:12.620 Pranav Narahari: Yeah.
371 00:29:13.820 ⇒ 00:29:16.439 Samuel Roberts: Because then, like, the whole history and everything has to, you know…
372 00:29:17.120 ⇒ 00:29:22.579 Samuel Roberts: It’s probably only MCP calls that we’re actually needed, because, like, I don’t think we’re gonna be hitting that with the, like, content window of the chat.
373 00:29:22.820 ⇒ 00:29:24.789 Samuel Roberts: Probably just when data comes back, actually.
374 00:29:24.790 ⇒ 00:29:27.999 Pranav Narahari: Yeah, the chat isn’t gonna be much, it’s just gonna be on…
375 00:29:28.700 ⇒ 00:29:32.929 Pranav Narahari: My only things with Claude’s skills, I know that they’re kind of,
376 00:29:32.930 ⇒ 00:29:39.079 Samuel Roberts: Oh, you’re right, you’re right. You’re right. Let me… let me… hold on, I was just looking that up real quick. Let’s see if there’s anything else here.
377 00:29:40.040 ⇒ 00:29:43.970 Samuel Roberts: Skills… First… oh, man, I lost my window, did I close it?
378 00:29:54.380 ⇒ 00:29:56.620 Samuel Roberts: Yeah, it’s also, like, tool calls and stuff.
379 00:29:57.240 ⇒ 00:30:03.630 Samuel Roberts: Because, like, the skills actually, like, call the code execution tool?
380 00:30:09.510 ⇒ 00:30:12.599 Samuel Roberts: And we can, yeah, we could also, like, give them a flip if they want to really start
381 00:30:12.730 ⇒ 00:30:13.819 Samuel Roberts: Dialing it in.
382 00:30:14.180 ⇒ 00:30:17.739 Samuel Roberts: like, this is using Haiku, this is using Sonic, kind of thing, I don’t know.
383 00:30:23.680 ⇒ 00:30:27.539 Samuel Roberts: Well, it’s… can the skills be used on all of them, is my question here.
384 00:30:33.890 ⇒ 00:30:39.569 Samuel Roberts: Alright, whatever, I’ll keep that. I would just say, yeah, give them that, like, explanation, and then we can dig into this later if they want us to.
385 00:30:47.520 ⇒ 00:30:48.469 Samuel Roberts: What else?
386 00:30:49.520 ⇒ 00:30:52.549 Casie Aviles: Yeah, I think we’re at time already, but…
387 00:30:53.030 ⇒ 00:30:55.180 Samuel Roberts: Oh, sorry, yeah, okay, we did get a little later on there.
388 00:30:55.180 ⇒ 00:30:56.289 Casie Aviles: Yeah, no problem.
389 00:30:56.290 ⇒ 00:30:57.220 Samuel Roberts: Apologies.
390 00:30:58.160 ⇒ 00:31:04.370 Casie Aviles: Yeah, so I think we’re… we’re… I think we’re clear, right? We should be clear on what… what…
391 00:31:06.090 ⇒ 00:31:10.599 Samuel Roberts: Yeah, let’s just look at the board real quick and make sure we’re… okay, Shopify data ingestion…
392 00:31:11.330 ⇒ 00:31:19.180 Samuel Roberts: In progress, forecasting progress, yeah. Yeah, I mean, I think we’re pretty, like, on track for Friday, at this point.
393 00:31:20.640 ⇒ 00:31:22.579 Samuel Roberts: Okay, cool. I’m trying to think of anything else.
394 00:31:27.140 ⇒ 00:31:30.080 Casie Aviles: Oh, yeah, I actually had a question.
395 00:31:30.080 ⇒ 00:31:30.670 Samuel Roberts: Yeah.
396 00:31:30.940 ⇒ 00:31:34.660 Casie Aviles: For the sources, so I know we did Shopify.
397 00:31:34.830 ⇒ 00:31:40.039 Casie Aviles: Should, should we also be doing… I think I wrote somewhere here.
398 00:31:41.280 ⇒ 00:31:47.070 Casie Aviles: Let me see… Meta and Clayview, or… Are those optional?
399 00:31:48.340 ⇒ 00:31:50.580 Samuel Roberts: Long-term, they’re not optional.
400 00:31:50.890 ⇒ 00:31:54.539 Samuel Roberts: Like, they’re gonna want that the same way we’re getting Shopify.
401 00:31:54.750 ⇒ 00:31:59.179 Samuel Roberts: In fact, they’re going to want to be doing all kinds of, like, data joins and stuff with it too, probably, so…
402 00:32:01.740 ⇒ 00:32:03.830 Samuel Roberts: It might be worth investigating that.
403 00:32:03.950 ⇒ 00:32:05.450 Samuel Roberts: Just to make sure that they work.
404 00:32:05.570 ⇒ 00:32:07.410 Samuel Roberts: Even if it’s not for the POC?
405 00:32:08.710 ⇒ 00:32:12.859 Casie Aviles: Okay, but this, this, this shouldn’t be required for the POC already?
406 00:32:14.540 ⇒ 00:32:16.080 Samuel Roberts: I don’t do it now.
407 00:32:16.540 ⇒ 00:32:17.210 Casie Aviles: Okay.
408 00:32:17.900 ⇒ 00:32:22.800 Samuel Roberts: I mean, I think… I think, like, that’s…
409 00:32:23.180 ⇒ 00:32:24.989 Samuel Roberts: Can you open up the Gantt real quick?
410 00:32:25.540 ⇒ 00:32:26.420 Casie Aviles: Okay, sure.
411 00:32:28.380 ⇒ 00:32:31.739 Samuel Roberts: 12-month forecast below the year. I mean, by the MVP, are we…
412 00:32:32.400 ⇒ 00:32:33.949 Samuel Roberts: Gonna need that stuff, you think?
413 00:32:34.240 ⇒ 00:32:38.409 Samuel Roberts: What is, what is… I may have to look at the actual PRD and see what it says.
414 00:32:39.540 ⇒ 00:32:44.790 Samuel Roberts: But my thought is just, like, if we do a quick check this week and make sure, that gives us more time to, like, verify.
415 00:32:44.940 ⇒ 00:32:47.590 Samuel Roberts: But we have another 2 weeks, so I’m not too worried.
416 00:32:48.470 ⇒ 00:32:54.410 Casie Aviles: Okay, cool. Yeah, because I… as far as I know, there are connections available.
417 00:32:54.950 ⇒ 00:32:57.930 Casie Aviles: or Facebook ads, or meta ads.
418 00:32:58.270 ⇒ 00:33:00.679 Samuel Roberts: Okay. I haven’t checked Clayview, but…
419 00:33:00.810 ⇒ 00:33:06.220 Casie Aviles: It’s… yeah, it could be there, because… Okay, cool.
420 00:33:06.220 ⇒ 00:33:10.410 Samuel Roberts: Yeah, I would say just double-check that and just post to the channel or something.
421 00:33:10.680 ⇒ 00:33:12.460 Samuel Roberts: The internal one, just let us know.
422 00:33:13.930 ⇒ 00:33:18.110 Samuel Roberts: But yeah, for POC, it’s just Shopify data, and the basic warehouse setup.
423 00:33:22.510 ⇒ 00:33:25.770 Casie Aviles: Okay, yeah, I think, I think that’s about it,
424 00:33:26.620 ⇒ 00:33:35.099 Casie Aviles: For these, do you guys need help with the… is there anything you want me to take on for these, ad hoc?
425 00:33:35.650 ⇒ 00:33:41.400 Samuel Roberts: Those, I would just… if you want to do a spike in one and Pranav does a spike in the other, or…
426 00:33:43.130 ⇒ 00:33:46.289 Samuel Roberts: No, I actually kind of want pronounced to do spikes on…
427 00:33:46.870 ⇒ 00:33:49.530 Samuel Roberts: Both, since he’s been talking with them about them?
428 00:33:50.340 ⇒ 00:33:53.069 Casie Aviles: But, Pranav, as part of that.
429 00:33:53.270 ⇒ 00:33:59.339 Samuel Roberts: Think about the… process we want to put in place for this, and, like, the standard, like.
430 00:34:00.000 ⇒ 00:34:02.650 Samuel Roberts: Standard operating procedure for when we get an ad hoc thing.
431 00:34:03.070 ⇒ 00:34:04.479 Samuel Roberts: That way, when there’s coming…
432 00:34:04.480 ⇒ 00:34:05.440 Pranav Narahari: So, I already have them.
433 00:34:05.440 ⇒ 00:34:05.850 Samuel Roberts: We can.
434 00:34:05.850 ⇒ 00:34:10.940 Pranav Narahari: with, like, the CSOs to talk about that. Yeah, so I think…
435 00:34:11.139 ⇒ 00:34:17.049 Pranav Narahari: The… the best… yeah, once we have that call, like, later this week,
436 00:34:17.480 ⇒ 00:34:20.009 Pranav Narahari: We should have a more, like, formalized way of.
437 00:34:20.010 ⇒ 00:34:23.769 Samuel Roberts: Oh, yeah, no, no, I mean, I’m thinking… you’re right, sorry, I should have been more clear.
438 00:34:23.880 ⇒ 00:34:29.599 Samuel Roberts: There’s the, like, how we handle it with the client, if there’s, like, You know, additional things…
439 00:34:29.800 ⇒ 00:34:36.389 Samuel Roberts: That are beyond scope. I’m thinking the, like, tech analysis of it in terms of, like, what will it take?
440 00:34:37.230 ⇒ 00:34:38.650 Pranav Narahari: Gotcha. Okay.
441 00:34:38.830 ⇒ 00:34:57.669 Samuel Roberts: Like, right now, I think you’re the one best positioned to do those spikes, just because you’ve been chatting with them and you’ve taken a look already, but I would like to get to a point where, like, any one of the three of us, or whoever else may be on the team in the future, there’s just a document that’s like, make sure these are in the PR, make sure they provided everything we need, and then do a quick, like.
442 00:34:58.020 ⇒ 00:35:00.790 Samuel Roberts: Analysis of how heavy a lift it is.
443 00:35:00.910 ⇒ 00:35:03.000 Pranav Narahari: Yeah, good call out. Yep.
444 00:35:03.400 ⇒ 00:35:08.699 Samuel Roberts: So, yeah, for those two, I would say definitely print off for now. You stick with them, but long term.
445 00:35:08.990 ⇒ 00:35:11.160 Samuel Roberts: Ad hoc should be able to be handled by any of us.
446 00:35:12.350 ⇒ 00:35:14.370 Samuel Roberts: At least through, like, the spike, I should say.
447 00:35:16.810 ⇒ 00:35:18.720 Samuel Roberts: Does that make sense? Does that sound reasonable?
448 00:35:19.100 ⇒ 00:35:20.700 Pranav Narahari: Yeah. Okay. Definitely.
449 00:35:21.630 ⇒ 00:35:22.310 Samuel Roberts: Cool, cool.
450 00:35:24.410 ⇒ 00:35:25.090 Casie Aviles: Alright.
451 00:35:25.550 ⇒ 00:35:29.469 Casie Aviles: Okay, cool. I think that’s, that’s all, really.
452 00:35:30.160 ⇒ 00:35:30.730 Samuel Roberts: Okay.
453 00:35:30.870 ⇒ 00:35:31.600 Samuel Roberts: Yeah.
454 00:35:31.770 ⇒ 00:35:32.350 Samuel Roberts: Thank you.
455 00:35:32.350 ⇒ 00:35:36.220 Casie Aviles: Yeah, I’ll just lock you guys if I have any other questions.
456 00:35:36.930 ⇒ 00:35:38.370 Samuel Roberts: Okay, sounds good.
457 00:35:40.370 ⇒ 00:35:41.750 Samuel Roberts: Yeah, I think we’re good then.
458 00:35:42.770 ⇒ 00:35:44.179 Casie Aviles: Okay, thank you.
459 00:35:44.560 ⇒ 00:35:45.850 Samuel Roberts: Alright, yeah, thanks, Casey.
460 00:35:46.500 ⇒ 00:35:47.130 Samuel Roberts: See yas.
461 00:35:47.130 ⇒ 00:35:47.800 Pranav Narahari: See you guys.