Meeting Title: Brainforge Data Warehousing Project Check-in Date: 2026-01-23 Meeting participants: Bobby Palmieri, zacfromson, Pranav Narahari, Samuel Roberts, Casie Aviles
WEBVTT
1 00:00:48.740 ⇒ 00:00:49.380 zacfromson: Yup.
2 00:00:49.870 ⇒ 00:00:50.750 Bobby Palmieri: Yo.
3 00:00:52.210 ⇒ 00:00:54.750 Bobby Palmieri: Can’t have you in that room later today.
4 00:00:54.750 ⇒ 00:00:56.479 zacfromson: Yeah, I’ll get out.
5 00:00:56.700 ⇒ 00:00:58.409 Bobby Palmieri: Put you in the living room.
6 00:00:58.690 ⇒ 00:01:00.780 zacfromson: Too big for a New York apartment, huh?
7 00:01:00.780 ⇒ 00:01:04.509 Bobby Palmieri: Yeah, and that window is, you know, there’s… yeah.
8 00:01:05.050 ⇒ 00:01:06.619 zacfromson: Can’t see, it’s glaring.
9 00:01:09.010 ⇒ 00:01:10.590 Bobby Palmieri: It’s up for now? What’s up?
10 00:01:10.590 ⇒ 00:01:11.510 Pranav Narahari: Gus.
11 00:01:11.510 ⇒ 00:01:15.640 Bobby Palmieri: Remove your location, since I talked to you 8 minutes ago?
12 00:01:18.810 ⇒ 00:01:23.530 Pranav Narahari: No, well, I am in a new location today, so Uten got a WeWork, and so…
13 00:01:23.530 ⇒ 00:01:24.340 Bobby Palmieri: submit today.
14 00:01:24.740 ⇒ 00:01:29.490 Pranav Narahari: Yeah, he’s in a table just, like, 20 feet from me, with some other people from Brainforge.
15 00:01:29.930 ⇒ 00:01:36.499 Bobby Palmieri: Should we, expect no productivity next week? Is Austin gonna shut down?
16 00:01:37.240 ⇒ 00:01:40.850 Pranav Narahari: I think by Monday we should be good, so just no productivity this weekend.
17 00:01:41.050 ⇒ 00:01:43.499 Bobby Palmieri: I’ll slide you, John.
18 00:01:44.120 ⇒ 00:01:46.750 Pranav Narahari: Sounds good, sounds good.
19 00:01:47.780 ⇒ 00:01:49.360 Pranav Narahari: Zach, how are you doing?
20 00:01:49.360 ⇒ 00:01:52.439 zacfromson: Doing pretty well here. How about you?
21 00:01:53.040 ⇒ 00:01:58.089 Pranav Narahari: Pretty good, pretty good, for now, while I’m still not… my blood is still flowing and not frozen.
22 00:01:58.090 ⇒ 00:01:58.940 zacfromson: Yeah.
23 00:01:58.940 ⇒ 00:02:04.599 Bobby Palmieri: What, what’s the timing on that? Is it… is it tonight, or Saturday and the weekend?
24 00:02:05.290 ⇒ 00:02:07.430 Pranav Narahari: I believe it starts tonight, like…
25 00:02:07.430 ⇒ 00:02:11.209 Bobby Palmieri: Yeah. Like, around, like, 5PM. Like, end of day, basically.
26 00:02:11.880 ⇒ 00:02:12.550 Bobby Palmieri: Got it.
27 00:02:12.910 ⇒ 00:02:23.559 Pranav Narahari: Yeah, yeah, tomorrow’s the worst of it, and then Sunday… actually, tomorrow’s the worst. Tomorrow’s pretty bad, but then also Monday, I just looked at it, and it looks pretty bad as of right now, so…
28 00:02:23.810 ⇒ 00:02:30.779 Pranav Narahari: we’ll see what happens. I feel like, in Texas, things just kind of get blown out of proportion. Like, it’s actually not gonna be that bad, but…
29 00:02:31.320 ⇒ 00:02:33.000 Pranav Narahari: We’ll see. We’ll see.
30 00:02:33.610 ⇒ 00:02:34.860 Bobby Palmieri: We dive on in.
31 00:02:35.450 ⇒ 00:02:40.379 Pranav Narahari: Let’s do it. Cool, yeah, Sam, Casey, y’all are here, perfect.
32 00:02:41.540 ⇒ 00:02:47.100 Pranav Narahari: Yeah, so I wrote up a quick agenda for what to do… what to go over today.
33 00:02:48.110 ⇒ 00:03:07.679 Pranav Narahari: And… yeah, so I kind of want to start things off with, data warehousing. And so, Sam and I, and Casey, talked to some other people, like, at Brainforge about that, kind of, like, what’s worked with them in the past, and I feel like we have a pretty, like, clear path forward on that. Sam, do you want to kind of dive into that?
34 00:03:08.030 ⇒ 00:03:09.600 Samuel Roberts: Yeah, so,
35 00:03:09.890 ⇒ 00:03:24.359 Samuel Roberts: Couple things. One, I was able to dump data from Shopify pretty easily, which was nice. Shopify’s API is really nice for that. So that was for testing purposes. We were able to get some stuff out, just get, like, the last two years of stuff.
36 00:03:24.470 ⇒ 00:03:30.159 Samuel Roberts: For future, we’re planning to dump everything into a DuckDB,
37 00:03:30.160 ⇒ 00:03:45.809 Samuel Roberts: instance. So DuckDB is the kind of technology… Mother Duck is the kind of cloud version of that. I don’t know if you guys have heard any of these before. Yeah, I mean, I’m new to some of this too, so, because I know, like, Postgres is, you know, for more…
38 00:03:45.820 ⇒ 00:03:51.730 Samuel Roberts: app data kind of thing, and duckDB is good for more analytic data, but it’s very similar in terms of, like.
39 00:03:51.780 ⇒ 00:04:07.459 Samuel Roberts: basic SQL stuff. For moving stuff in, we were looking a bit, there’s a ton of different ETLs out there, and they’re all different kind of price points and, you know, features, and certain things are probably a little more than we need right now, so AirByte looks pretty good.
40 00:04:07.460 ⇒ 00:04:17.259 Samuel Roberts: The question becomes AirByte Cloud versus running it because it is open source. And then the other questions we need to still answer are, you know, it’s gonna have…
41 00:04:17.370 ⇒ 00:04:35.070 Samuel Roberts: 70-something different Shopify stores to sync, right? So making sure that it handles all of these API keys well, syncs them all properly to the DB, because the other side of it is, we have these four sources. I guess a question I have is, will there be other sources in the future? Are these kind of the main ones, like.
42 00:04:35.230 ⇒ 00:04:40.819 zacfromson: there’ll be more. I think, like, there’s a lot of Shopify apps that we’re gonna, like, even want to tie into. There’s, like.
43 00:04:40.820 ⇒ 00:04:48.030 Samuel Roberts: Sure, okay. That’s what I would… and there’s all these other third-party apps, and Gorgeous, we’re going to want to tap into for most of them. Of course, okay.
44 00:04:48.030 ⇒ 00:04:59.000 zacfromson: I think there’s a laundry list of other things that we’ll want to go through. I think we’ll eventually want to tap on Amazon, Applovin, TikTok, Snapchat. So, yeah, we’re just probably getting started. These are the core pieces, but I think it’s.
45 00:04:59.000 ⇒ 00:04:59.410 Samuel Roberts: That’s what I think.
46 00:04:59.410 ⇒ 00:05:00.840 zacfromson: The build we’re gonna get.
47 00:05:00.840 ⇒ 00:05:17.810 Samuel Roberts: And that’s kind of the way the MCPs are set up right now, is you can… we can more easily add MCPs, and it’s reading from the database and everything, so we can update that without having to, like, you know, restructure things. I assumed that was kind of the case for this, too, so if there’s a list of, like, those types of things, it might be good to check against Airbyte’s.
48 00:05:18.040 ⇒ 00:05:27.550 Bobby Palmieri: Yeah, we’ve… we’ve checked against AirByte, Zach’s talked to two other people, and I talked to two other agencies, and they’re all using AirByte, and have not run into any… Okay, that’s what I want to do.
49 00:05:27.550 ⇒ 00:05:38.769 zacfromson: How we ended up with you guys, Sawyers, like, building some products, like, he’s basically building, like, some products, like, in the similar D2C space, and, like, I was, like, the first beta tester on his product, and…
50 00:05:38.770 ⇒ 00:05:39.320 Samuel Roberts: Oh, nice, okay.
51 00:05:39.320 ⇒ 00:05:50.170 zacfromson: They’re using AirByte, like, it was, like, loading up. Do you guys use AirByte? He’s like, yep, it’s the way to go, so… Okay. That icon before this as well, and they were using AirByte there, and all his engineers, so that’s, like, the way to go, so…
52 00:05:50.170 ⇒ 00:05:50.850 Bobby Palmieri: My.
53 00:05:50.850 ⇒ 00:05:53.440 zacfromson: I don’t know much, but it’s just recommendations, Scott, Bobby.
54 00:05:53.810 ⇒ 00:05:57.440 Bobby Palmieri: Yeah, my understanding from, like, everyone that I’ve talked to.
55 00:05:57.580 ⇒ 00:06:07.079 Bobby Palmieri: that’s building in a similar space, and, you know, everyone has different requirements, but they’re using… either using Fivetran or AirByte, but AirByte seems to be more…
56 00:06:07.970 ⇒ 00:06:19.540 Bobby Palmieri: built for what we want to do, and they have every connection that I’ve looked for for us, future… or, like, current and future state, so I think we can just put AirByte in the done column and move forward.
57 00:06:19.540 ⇒ 00:06:23.730 Samuel Roberts: That’s what I wanted to make sure, because, yeah, 5chan, I know is, like, kind of…
58 00:06:24.270 ⇒ 00:06:36.679 Samuel Roberts: one of the standard ones, but it’s also, like, way up there in price per… like, and so I don’t know how that scales as well, in terms of, like, that sort of stuff, so… Okay. Do you know if people are using AirByte Cloud, or are they hosting it? Does that mean anything? Okay.
59 00:06:36.680 ⇒ 00:06:38.429 zacfromson: I don’t know, I could ask Sawyer what they’re doing in there.
60 00:06:38.430 ⇒ 00:06:46.110 Samuel Roberts: Okay, because I’m just… that’s the other thing I want to make sure we understand, like, if we’re… you know, some people are just, like, ingesting their own stuff, and I want to make sure that if we’re ingesting
61 00:06:46.140 ⇒ 00:06:58.650 Samuel Roberts: tons of them. If we go with the cloud, it’s not gonna incur, like, you know, scaling it that way, versus us managing a little more infrastructure might be worth it at that point, but, it’s all open source, so it’s nice to be able to at least see the docs that way.
62 00:06:59.820 ⇒ 00:07:07.899 Samuel Roberts: So yeah, I think that’s… that’s fine. That’s exactly… that was the one kind of, like, I didn’t know what other sources to confirm on AirByte, but if you guys have already jumped that, we’re good then.
63 00:07:08.190 ⇒ 00:07:14.979 Samuel Roberts: So yeah, then, one of the other things we were talking about was things like Snowflake, which is another big, warehouse, but that’s…
64 00:07:15.310 ⇒ 00:07:28.930 Samuel Roberts: probably overkill for what we’re doing here, so, like, you know, the way I’m thinking about it right now is, like, the data’s coming in, it’s gonna sit in the warehouse, we’ll run some models on that, and then, you know, the forecasting tool and other tools will kind of sit on top of that. Snowflake is for, like.
65 00:07:29.150 ⇒ 00:07:42.450 Samuel Roberts: people running all kinds of queries and doing all kinds of stuff and changing things all the time, and so the way we’re setting it up is, like, most of that stuff is pretty well… gonna be well established. You know, you’re not gonna be sitting there running SQL queries against it, it’s gonna all be part of the app.
66 00:07:42.470 ⇒ 00:07:49.709 Samuel Roberts: So I think we’re pretty well set up with DuckDB in here by then. So that’s kind of the plan moving forward.
67 00:07:50.670 ⇒ 00:07:51.360 Pranav Narahari: You know.
68 00:07:51.360 ⇒ 00:08:03.710 Samuel Roberts: Long term. In the immediate future, we’re gonna probably take that Shopify data that I already got into a DuckDB file, and be able to build the forecasting tool, proof of concept on top of that, and make sure everything goes the way we expect, so…
69 00:08:04.080 ⇒ 00:08:12.370 Pranav Narahari: Yeah, Sam, on Snowflake, that’s kind of if you wanted to have, like, a bunch of different dashboards pulling in data in different ways, right? Like, various queries?
70 00:08:12.370 ⇒ 00:08:14.829 Samuel Roberts: Yeah, yeah, and so, like, you know, they use.
71 00:08:14.830 ⇒ 00:08:16.019 Pranav Narahari: It’s not really…
72 00:08:16.020 ⇒ 00:08:22.330 Samuel Roberts: Exactly, exactly. Like, it’s a warehouse the same way it’s in the warehouse, but, like, they have different feature sets that we just don’t need for this project.
73 00:08:22.860 ⇒ 00:08:23.830 Pranav Narahari: Totally. Okay.
74 00:08:24.380 ⇒ 00:08:30.380 Pranav Narahari: Cool, so I think we’re all aligned there. Yeah, we already did the Shopify, like.
75 00:08:30.470 ⇒ 00:08:45.980 Pranav Narahari: especially for the POC, just, like, the last two years’ data, we were able to download that. I think you were saying, like, with Shopify, they had a really, like, neat agent on the side for us to, like, do that pretty… or at least, like, help supplement that, like, pretty easily, so that’s a good tool that we found out for the future.
76 00:08:46.120 ⇒ 00:08:57.739 Pranav Narahari: Yeah, so that’s good to know. One other piece of big news is production is now live. I’m gonna send you guys this link, and you guys can poke around.
77 00:08:59.890 ⇒ 00:09:07.490 Pranav Narahari: But, yeah, Sam put in some, like, good work this week to kind of set up that entire, like, deploy process, and…
78 00:09:08.070 ⇒ 00:09:11.609 Pranav Narahari: Sorry, my Wi-Fi has been really bad. Can you guys still hear me?
79 00:09:12.450 ⇒ 00:09:13.449 Bobby Palmieri: Yeah, yeah, we can hear you.
80 00:09:13.450 ⇒ 00:09:15.919 Pranav Narahari: Okay, perfect. I’m just gonna turn off my video.
81 00:09:15.920 ⇒ 00:09:18.160 Bobby Palmieri: Yeah, your audio’s been great, but you keep freezing.
82 00:09:18.160 ⇒ 00:09:18.720 Samuel Roberts: Yeah.
83 00:09:18.720 ⇒ 00:09:34.059 Pranav Narahari: Okay, yeah, I’m just gonna turn off my video, hopefully that helps. But, yeah, so production’s now live, you guys can poke around in there. One thing that we’re just waiting on is that CSV to just see the… the brands and all of their MCP connections.
84 00:09:34.060 ⇒ 00:09:35.679 Bobby Palmieri: I’m getting that this weekend.
85 00:09:36.510 ⇒ 00:09:42.049 Samuel Roberts: Yeah, even if it’s just, like, a few, and you want to do it, kind of, in stages, or however long it takes to put that together, you know.
86 00:09:42.050 ⇒ 00:09:49.320 zacfromson: Yeah, I wanted to link with you. I don’t have access to everything, Bobby, and that’s why, like, I’m not positive in the shop I set up, you’re the one with access to Anthropic, so I can pull…
87 00:09:49.320 ⇒ 00:09:52.170 Bobby Palmieri: We’ll have Anissa do all the API keys.
88 00:09:52.750 ⇒ 00:10:08.280 zacfromson: Yeah, I have a sheet set up for us to use. I just took, like, Sam’s in, and I, like, set something up and drive it. I just haven’t started filling it out. So, I mean, you can think about it this weekend and build a plan to get that built out. Or one of our VAs is just out this week on vacation, so I’m just waiting for her to get back to do, you know, all set these counts.
89 00:10:08.280 ⇒ 00:10:13.450 Samuel Roberts: Does that CSV have the Gemini key as well? That might have been before that was… so you can easily just add another one.
90 00:10:13.450 ⇒ 00:10:15.229 zacfromson: No, but we’ll tag on there.
91 00:10:15.230 ⇒ 00:10:16.210 Samuel Roberts: Yeah, that’s fine.
92 00:10:16.210 ⇒ 00:10:18.970 zacfromson: connections, make sure everything’s in there, and
93 00:10:19.320 ⇒ 00:10:26.780 zacfromson: We’ll go from there. One other note, too… yeah, that’s good. The only thing is that the brand description, I think, is, like, in… in there. I think it’s, like, mandatory.
94 00:10:26.910 ⇒ 00:10:32.939 Samuel Roberts: Yeah, I noticed that when I was testing creating a test brand, we can… we can flip that off for mandatory, so it can just be the name, and then make the…
95 00:10:32.940 ⇒ 00:10:35.879 zacfromson: That’s a necessary field, so that’d be helpful.
96 00:10:35.880 ⇒ 00:10:44.460 Samuel Roberts: Yeah, I’ll definitely, I’ll definitely flip that. I mean, we can always create, you know, the UI has its restrictions, and like, but if you don’t include it in the CSV, it shouldn’t be a problem creating them, so…
97 00:10:44.540 ⇒ 00:10:49.240 Samuel Roberts: Cool. Yep. You can even update that. We’ll update that in the UI anyway, because I noticed that when I was creating.
98 00:10:49.240 ⇒ 00:11:04.729 Samuel Roberts: There’s a couple other little things, even just, like, clicking around testing in prod, little quality of life things that… that I saw, that UTom saw, and we have a little bit of a list going. I imagine they’ll be similar to some of the things you’ll see as you start testing it out. Just, like, little, like, when you click settings, it jumps to the…
99 00:11:04.780 ⇒ 00:11:13.129 Samuel Roberts: the settings instead of just doing the drop-down. Part of that came from, like, redoing the URLs, and there needed to be a URL for just going to settings, for example.
100 00:11:13.240 ⇒ 00:11:18.240 Samuel Roberts: Little, little, like, quality of life things that we can, we can improve and push out, you know, kind of like…
101 00:11:18.600 ⇒ 00:11:25.170 Samuel Roberts: like, a small patch of fixes to production without it being, like, a big… a big thing. So I know you guys mentioned that you were
102 00:11:25.270 ⇒ 00:11:30.490 Samuel Roberts: putting together, like, a list of that sort of stuff with priorities and stuff. So, we have a few things running as well, so, you know…
103 00:11:30.680 ⇒ 00:11:31.889 Samuel Roberts: We can add to that.
104 00:11:31.980 ⇒ 00:11:32.860 Bobby Palmieri: Awesome.
105 00:11:33.120 ⇒ 00:11:35.590 Bobby Palmieri: I had another thought on that, and I don’t know where it was now.
106 00:11:36.310 ⇒ 00:11:37.200 Samuel Roberts: That’s enough.
107 00:11:37.540 ⇒ 00:11:43.290 Pranav Narahari: One thing I wanted to mention with, like, production and staging is how we kind of want to view these two environments now.
108 00:11:43.290 ⇒ 00:11:43.680 Samuel Roberts: That’s a bit.
109 00:11:43.680 ⇒ 00:12:00.819 Pranav Narahari: different. So I think staging is all you guys had before, so you guys are just doing everything in there. Now, we kind of want to use staging as part of our, like, final stage of our deploy process. So, like, once you guys give us the clear on a certain feature in our staging environment, if needed, you know, some things, maybe if they’re backend changes.
110 00:12:00.820 ⇒ 00:12:07.400 Pranav Narahari: You guys might not need to review, but for certain things, we’ll ask you guys to review it, maybe just do, like, some…
111 00:12:07.900 ⇒ 00:12:16.090 Pranav Narahari: like, QA testing, that shouldn’t take too long, and then that way, when they’re in production, we don’t feel like we’re doing, like, a bunch of, like, patches last minute.
112 00:12:16.460 ⇒ 00:12:17.260 Bobby Palmieri: Yeah.
113 00:12:17.310 ⇒ 00:12:20.589 Pranav Narahari: And so, that’s kind of how we’re gonna start viewing staging, so…
114 00:12:20.840 ⇒ 00:12:23.279 Pranav Narahari: Production will probably have, like, a ton of users.
115 00:12:23.800 ⇒ 00:12:25.379 Bobby Palmieri: We’ll move to, like, the dev…
116 00:12:26.270 ⇒ 00:12:28.699 Bobby Palmieri: Server to, like, test new features, is that right?
117 00:12:30.460 ⇒ 00:12:31.770 Pranav Narahari: Yeah.
118 00:12:31.770 ⇒ 00:12:33.110 Samuel Roberts: Go ahead, go ahead, Ron.
119 00:12:33.740 ⇒ 00:12:48.780 Pranav Narahari: Yeah, so, well, so we have dev, staging, and prod. Dev is what we’re going to, like, set you guys up with for Cloud Code. So, like, if you guys want to create new features, like, that’s what it’s going to wire up to. Staging, though, is going to be, like.
120 00:12:49.080 ⇒ 00:12:55.359 Pranav Narahari: it should be, every once in a while, in sync with prod, and it’s just going to be, like, our, like, final, like.
121 00:12:55.360 ⇒ 00:12:58.049 Bobby Palmieri: testing before we move it to prod, yeah.
122 00:12:58.050 ⇒ 00:13:07.089 Pranav Narahari: Yeah, just another… another intermediary before, you know, we throw things in production. Yeah, anything else you wanted to add to that, Sam?
123 00:13:07.670 ⇒ 00:13:16.840 Samuel Roberts: Yeah, so, yes, think of production as the app, you know, get people in there, get people using it. Obviously, like, there’s still plenty of things that will be added and tested, you know.
124 00:13:16.840 ⇒ 00:13:18.200 Bobby Palmieri: I think you’re probably there.
125 00:13:18.380 ⇒ 00:13:20.220 Samuel Roberts: The more people that are in there, especially.
126 00:13:20.340 ⇒ 00:13:28.510 Samuel Roberts: But yeah, so we’ll make sure that database will be backed up regularly, like, treat that data kind of as sacred. The staging data is basically, like.
127 00:13:28.510 ⇒ 00:13:47.009 Samuel Roberts: we’ll probably pull down from prod to link that and make sure everything’s still working when we put new features and stuff. So don’t rely on any data in staging, don’t rely on any data in dev if anything’s important, and get it out of there, because it’s… it’s… we’re just treating that as, like, more volatile, basically, than production, which is the data that will be backed up and, you know, restorable and stuff like that.
128 00:13:48.230 ⇒ 00:13:49.760 Samuel Roberts: But yeah, that’s pretty much it.
129 00:13:50.340 ⇒ 00:13:56.670 Pranav Narahari: Exactly. Okay, cool. Also, so I think at the beginning of this week, we were just,
130 00:13:57.020 ⇒ 00:14:04.809 Pranav Narahari: we set you guys up with Claude Code, and then Bobby, you kind of created a PR for Claude Code as well.
131 00:14:05.190 ⇒ 00:14:11.620 Pranav Narahari: I kind of wanted to talk about just, like, what’s been working well for you guys, what you think could be improved there,
132 00:14:12.320 ⇒ 00:14:18.969 Pranav Narahari: any… any… I have some, like, you know, how I think we can improve that process, but I’m just wondering from you guys first, like…
133 00:14:19.120 ⇒ 00:14:20.420 Pranav Narahari: Any comments?
134 00:14:20.770 ⇒ 00:14:31.299 Bobby Palmieri: Yeah, I don’t have any, like, feedback yet. I think I ripped together, like, two features that we’re pretty excited about. The dev…
135 00:14:31.790 ⇒ 00:14:36.180 Bobby Palmieri: front end looked really solid, and, like, Claude was telling me that, like.
136 00:14:36.550 ⇒ 00:14:47.180 Bobby Palmieri: I don’t know, it was asking me, like, I need access to this or, like, that, and I was like, I’m not fucking touching that. So, I pushed it to you guys, it, like… I did have it, like, load some demo data so I could see…
137 00:14:47.430 ⇒ 00:15:00.030 Bobby Palmieri: that, but it… and it said, like, oh, once you connect to Railway and this and that, like, it will work with real data. So, I pushed those two PRs to GitHub, and with…
138 00:15:00.130 ⇒ 00:15:10.149 Bobby Palmieri: love to know, like, hey, was… did that work? And then kind of… and then I have, like, other questions of, like, okay, now I don’t actually know
139 00:15:10.340 ⇒ 00:15:24.479 Bobby Palmieri: how to, like, spin it up again and do it again, but, like, I wanted… I just… I made those two, I kind of spent, like, I don’t know, four to six hours kind of vibe coding two features that I actually think came out pretty good,
140 00:15:25.160 ⇒ 00:15:27.090 Bobby Palmieri: Providing that they work.
141 00:15:27.900 ⇒ 00:15:31.880 Bobby Palmieri: And then I stopped there, because I was like, I’m not gonna go build…
142 00:15:32.090 ⇒ 00:15:41.349 Bobby Palmieri: anything else for you guys to be like, hey, this is spaghetti, and we can’t use any of it. So, I can walk through, kind of, like, what I built, and…
143 00:15:42.060 ⇒ 00:15:44.550 Bobby Palmieri: Or, like, if you guys have already seen it, but…
144 00:15:45.110 ⇒ 00:15:57.840 Pranav Narahari: let’s… yeah, let’s actually go into that, too. One thing I just want to mention for, like, the two things that you mentioned, so, like, yeah, we did see those PRs come through to GitHub. Everything, like, end-to-end process worked perfect.
145 00:15:57.970 ⇒ 00:16:08.400 Pranav Narahari: The other thing that you mentioned that, at least I think it was kind of on the same vein of, like, what I was thinking could be improved is, like, okay, now you built out a feature, say you want to work on something else.
146 00:16:08.830 ⇒ 00:16:15.559 Pranav Narahari: you don’t want to build that actually in the same branch, because then you’re just coupling code that doesn’t necessarily need to be coupled.
147 00:16:15.620 ⇒ 00:16:30.040 Pranav Narahari: And so, yeah, we’re gonna… we have a process, like, we have a design in mind, and I’ll create, like, a loom for you guys to show you, like, okay, how you can prompt Claude Code to just basically create a new branch off of dev,
148 00:16:30.040 ⇒ 00:16:38.050 Pranav Narahari: and basically… well, actually, it’ll be off of your Vibecoding environments, and then you can just, like, vibecode a new feature.
149 00:16:38.380 ⇒ 00:16:38.970 Bobby Palmieri: Yeah.
150 00:16:39.170 ⇒ 00:16:42.649 Pranav Narahari: So it should be pretty, pretty straightforward there.
151 00:16:43.060 ⇒ 00:16:49.629 Pranav Narahari: But yeah, those are the two things I want to mention. Yeah, if you want to show what you made, and then we can kind of talk a little bit about, like.
152 00:16:50.980 ⇒ 00:16:55.289 Pranav Narahari: what, Sam and I need to do in terms of, like, backend improvements, like…
153 00:16:55.290 ⇒ 00:17:07.039 Samuel Roberts: That’s what I was going to mention as well, you know, there’s… there’s a couple ways to go with the backend things. One is, you know, if this is working well, and, you know, Claude’s doing what you want, like, we could let you
154 00:17:07.210 ⇒ 00:17:17.240 Samuel Roberts: vibe on the back end a little bit, and see if Claude does what’s doing well. The other side of it is, if not, we probably need to know, like, what Claude thinks needs to happen next, based on those changes.
155 00:17:18.000 ⇒ 00:17:24.380 Samuel Roberts: Which you could probably get Claude to, like, spit out a Markdown file for us that we could, you know, run with, because,
156 00:17:24.599 ⇒ 00:17:35.879 Samuel Roberts: you know, it might not be clear exactly, like, what endpoints it needs, or what logic it needs, if we’re just looking at the PR, necessarily. So making sure to, like, include that kind of stuff from Claude would be a huge help there.
157 00:17:36.290 ⇒ 00:17:42.650 Bobby Palmieri: So I don’t know what I need to do… this, like, localhost just says it’s redirecting, and it doesn’t have any of the demo stores.
158 00:17:43.100 ⇒ 00:17:46.649 Bobby Palmieri: Any thoughts on what I need to prompt Claude to do?
159 00:17:50.650 ⇒ 00:17:55.489 Pranav Narahari: Sometimes, if it just, like, has, like…
160 00:17:55.620 ⇒ 00:18:05.109 Pranav Narahari: the cookies stored from, like, staging or production, like, it does this, so you could try to, like, just delete your cookies,
161 00:18:05.980 ⇒ 00:18:14.219 Pranav Narahari: Yeah, sometimes restarting works too, so if you want to, like, yeah, if you want to go back to… yeah, you can just do, so do INCOG, too.
162 00:18:22.740 ⇒ 00:18:25.049 Bobby Palmieri: Let me just try and restart here.
163 00:18:26.460 ⇒ 00:18:27.769 Pranav Narahari: Yeah, we can try both.
164 00:18:29.730 ⇒ 00:18:30.400 Bobby Palmieri: Doesn’t seem…
165 00:18:37.230 ⇒ 00:18:39.159 Bobby Palmieri: Should I maybe start a new chat?
166 00:18:39.780 ⇒ 00:18:41.060 Bobby Palmieri: Oh, no, there it goes.
167 00:18:45.620 ⇒ 00:18:48.709 Pranav Narahari: What can happen, too, is, like, at the end of your, like.
168 00:18:49.110 ⇒ 00:18:57.479 Pranav Narahari: Whenever you’re vibe coding, like, when you start a server, if you don’t tell it to stop, it might… and you just go back to the same, like, thread.
169 00:18:57.800 ⇒ 00:18:58.140 Bobby Palmieri: Yeah.
170 00:18:58.140 ⇒ 00:19:06.209 Pranav Narahari: Like, if that thread just… if that, server just is, like, up for a long period of time, like, there could be some errors, and if you start another one.
171 00:19:06.500 ⇒ 00:19:11.049 Pranav Narahari: That’s where things can get a little bit funky, so you just kind of… need to…
172 00:19:11.560 ⇒ 00:19:21.120 Pranav Narahari: if here, like, it… you could probably… because I just saw, like, higher on the thread, I saw that it said, like, okay, development server’s already running, so if it’s running two apps at the same time.
173 00:19:21.580 ⇒ 00:19:26.529 Pranav Narahari: the ports can get mixed up, things can get kind of weird.
174 00:19:27.150 ⇒ 00:19:30.610 Pranav Narahari: But yeah, usually just, like, if you just tell it to, like, Stop.
175 00:19:30.880 ⇒ 00:19:33.410 Pranav Narahari: Oh, and I think it’s doing it right now.
176 00:19:34.220 ⇒ 00:19:35.219 Pranav Narahari: But, yeah, let’s…
177 00:19:35.220 ⇒ 00:19:37.069 Bobby Palmieri: Multiple sale processes.
178 00:19:37.510 ⇒ 00:19:42.450 Pranav Narahari: Yeah, yeah, yeah, that’s probably what happened. So yeah, okay, it’s cleaning that up. After this, it should be fine.
179 00:19:48.810 ⇒ 00:19:49.200 Bobby Palmieri: There we go.
180 00:19:49.200 ⇒ 00:19:49.810 Pranav Narahari: Perfect.
181 00:19:50.590 ⇒ 00:19:56.809 Bobby Palmieri: So… I created two things, so…
182 00:19:56.990 ⇒ 00:20:08.640 Bobby Palmieri: One is a meta ads performance dashboard, that pulls in, you know, a variety of different things.
183 00:20:09.270 ⇒ 00:20:15.330 Bobby Palmieri: You know, across different periods of time. And it says that it’s loading demo data.
184 00:20:16.220 ⇒ 00:20:20.319 Bobby Palmieri: But it, like, when you push it to the back end, it should be…
185 00:20:20.520 ⇒ 00:20:29.559 Bobby Palmieri: connected, and I’m fine to prompt it. I included some in my thread there of, like, what it said it needed. But supposedly this
186 00:20:29.630 ⇒ 00:20:41.440 Bobby Palmieri: should work… I mean, I guess it… Claude believes that it should work, so this is, like, feature number one to pull in from the Facebook API, like, these different breakdowns.
187 00:20:41.650 ⇒ 00:20:44.560 Samuel Roberts: Any questions there, or things that you guys want to see?
188 00:20:45.610 ⇒ 00:20:51.880 Pranav Narahari: No, this makes sense to me, like, all of these fields look very familiar as… since I was working on, like, the Slack reports.
189 00:20:51.880 ⇒ 00:20:52.450 Bobby Palmieri: Yup.
190 00:20:52.670 ⇒ 00:20:59.639 Pranav Narahari: They’re just more aggregated. Or actually, no, these even look like they’re for individual days, too, so… Yeah, Sam, this seems like.
191 00:21:00.340 ⇒ 00:21:02.669 Bobby Palmieri: Daily, weekly, monthly, quarterly.
192 00:21:02.760 ⇒ 00:21:03.450 Samuel Roberts: Gotcha.
193 00:21:04.000 ⇒ 00:21:14.179 Pranav Narahari: Yeah, and so this is, like, a perfect use case with the data warehouse, because right now, for the slag reports, for us to wire it up, we are doing, like, the calculations, like, ad hoc.
194 00:21:15.530 ⇒ 00:21:21.689 Pranav Narahari: And that is just, like, a lot of load, and, like, say there’s, like, you know, tons of scheduled reports.
195 00:21:22.160 ⇒ 00:21:25.759 Pranav Narahari: We’re just putting a lot of load on our servers.
196 00:21:25.940 ⇒ 00:21:33.970 Pranav Narahari: And so, yeah, with the data warehouse stuff, like, this is, like, a really good use case for… for this, basically. Like, it should be pretty much seamless.
197 00:21:34.080 ⇒ 00:21:37.380 Pranav Narahari: Yeah, Sam, anything else?
198 00:21:38.230 ⇒ 00:21:54.880 Samuel Roberts: Yeah, I mean, you hit exactly what I was thinking. Like, this may even work pushing, because it does have access to, like, you know, the meta keys and being able to fetch the data. The question becomes, you know, every time you click one of those, is it refreshing the data? You know, if it’s not being stored properly yet in the warehouse, like, how is that gonna… so, yeah, this…
199 00:21:54.880 ⇒ 00:22:01.029 Samuel Roberts: is gonna be really good once the warehouse is set up, because we’ll just be able to pull right from that, and the models that we have set up already.
200 00:22:01.820 ⇒ 00:22:04.960 Bobby Palmieri: Yeah, I think we chatted about this last week,
201 00:22:05.190 ⇒ 00:22:09.430 Bobby Palmieri: that, like, I think, Sam, you had mentioned that, like, it won’t be too much of a…
202 00:22:10.350 ⇒ 00:22:14.080 Bobby Palmieri: Rework to then, once we have the data warehouse, to pull in from that.
203 00:22:14.080 ⇒ 00:22:31.319 Samuel Roberts: Yeah, yeah, the idea is, like, we can fetch data on the fly, we’re just not storing it. So once we start storing it, it’ll be, you know, even easier to just pull from that source, and faster, and, you know, less load, all these things. But yeah, for testing purposes, like, this is… this is great. We can even try to spin this up on Railway and see what actually happens.
204 00:22:31.320 ⇒ 00:22:31.890 Bobby Palmieri: Oh.
205 00:22:32.830 ⇒ 00:22:45.320 Samuel Roberts: Yeah, that’s totally worth trying to see what it thinks it needs. Because, yeah, if it’s, you know, for… until we get the warehouse set up, if it’s just taking a minute to load the page, that’s, you know, a small price to pay for now, but we can at least keep moving.
206 00:22:45.550 ⇒ 00:22:51.340 Bobby Palmieri: The second one that I spun up was this email calendar. I think this may not…
207 00:22:52.600 ⇒ 00:23:04.940 Bobby Palmieri: generate demo data, I could prompt it to do it. But essentially, like, this is pulling in from Klaviyo, and this will be a good transition into Claude Skills.
208 00:23:04.940 ⇒ 00:23:14.829 Bobby Palmieri: But essentially, like, you can prompt, you know, whatever month you want, right? How many emails you want in the calendar, how many SMS you want.
209 00:23:15.230 ⇒ 00:23:21.370 Bobby Palmieri: Do you want to use historical context from, like, last month, 2 months, 3 months for… Klaviyo?
210 00:23:21.560 ⇒ 00:23:34.720 Bobby Palmieri: should you include, like, year-over-year context, and then be like, you know, I want… it’s like white font, but, you know, I want to include a Valentine’s Day
211 00:23:34.720 ⇒ 00:23:45.369 Bobby Palmieri: sale… whatever else. And then when I generate it, yeah, so it’s gonna fail to generate. But I think, actually…
212 00:23:56.000 ⇒ 00:24:03.370 Bobby Palmieri: It gave you, like, the description of what it needed here, but if I, like, scroll up…
213 00:24:06.180 ⇒ 00:24:10.939 Bobby Palmieri: on this… You can see the…
214 00:24:11.800 ⇒ 00:24:17.730 Bobby Palmieri: framework that I gave it, because it was actually giving me demo data, and then when I wanted to push, I was like, remove the demo data.
215 00:24:17.730 ⇒ 00:24:18.990 Samuel Roberts: Yeah.
216 00:24:20.790 ⇒ 00:24:29.360 Bobby Palmieri: I don’t know if it… We’ll go back far enough, but, I can show you, like, what… Wee…
217 00:24:30.540 ⇒ 00:24:39.130 Bobby Palmieri: Created… Somewhere… I also went through a…
218 00:24:39.790 ⇒ 00:24:43.709 Bobby Palmieri: I also, like, deleted everything that I did, and then I had to, like.
219 00:24:43.880 ⇒ 00:24:50.979 Bobby Palmieri: Chat with it for 2 hours to get it back, but, yeah, that happens sometimes.
220 00:24:51.230 ⇒ 00:24:52.420 Samuel Roberts: Not uncommon.
221 00:24:52.690 ⇒ 00:24:58.589 Bobby Palmieri: So I’ll find the screenshots of, like, the demo data, but essentially it generates, like, a,
222 00:25:01.160 ⇒ 00:25:04.200 Bobby Palmieri: Calendar… here it is, once this loads.
223 00:25:11.860 ⇒ 00:25:13.850 Bobby Palmieri: It’s right there, it’s just disappearing.
224 00:25:16.320 ⇒ 00:25:19.539 Bobby Palmieri: So it, like, creates a, like, a content calendar.
225 00:25:20.280 ⇒ 00:25:22.269 Bobby Palmieri: Creates, like, this timeline view.
226 00:25:22.400 ⇒ 00:25:24.139 Bobby Palmieri: Like, here are the emails.
227 00:25:25.250 ⇒ 00:25:30.839 Bobby Palmieri: And then you can generate a brief, which will do, like, copywriting for the email.
228 00:25:31.340 ⇒ 00:25:32.760 Samuel Roberts: We don’t know, yeah.
229 00:25:33.240 ⇒ 00:25:40.320 Bobby Palmieri: So… That’s… That’s probably the feature that we’re most, like, excited about getting.
230 00:25:40.830 ⇒ 00:25:54.110 Bobby Palmieri: live, but I don’t… again, I have no idea how far away this is from, like, production, but it would be good to know, because honestly, the Cloud Code setup that we have worked really well, and I was very happy with
231 00:25:54.660 ⇒ 00:25:57.980 Bobby Palmieri: what it told me and what I saw, I just don’t know if it was…
232 00:25:58.370 ⇒ 00:26:04.450 Bobby Palmieri: you know, I don’t know if it’s functional, once it has access to the MCP and stuff like that.
233 00:26:05.080 ⇒ 00:26:08.929 Pranav Narahari: Yeah, so for the email sender, like, or the email one.
234 00:26:09.140 ⇒ 00:26:15.340 Pranav Narahari: Sam, that seems like not really dependent on Data Warehouse, so we can probably start to rip that as soon as next week.
235 00:26:15.870 ⇒ 00:26:19.779 Bobby Palmieri: Yeah, it is dependent on Klaviyo MCP of, like, getting context.
236 00:26:19.780 ⇒ 00:26:20.300 Samuel Roberts: Yeah.
237 00:26:20.300 ⇒ 00:26:21.580 Bobby Palmieri: And then…
238 00:26:21.580 ⇒ 00:26:21.970 Pranav Narahari: Yeah.
239 00:26:21.970 ⇒ 00:26:25.669 Bobby Palmieri: also has a Claude skill that we want to use for…
240 00:26:25.670 ⇒ 00:26:26.090 Pranav Narahari: Gotcha.
241 00:26:26.190 ⇒ 00:26:33.880 Bobby Palmieri: calendar generator, but that’s just part of, like, the prompt, I guess, is, like, use this skill, as well.
242 00:26:34.940 ⇒ 00:26:36.030 Pranav Narahari: Gotcha.
243 00:26:36.870 ⇒ 00:26:39.319 Pranav Narahari: Okay. Cool.
244 00:26:39.770 ⇒ 00:26:53.249 Pranav Narahari: Yeah, one thing that, Sam and I were also talking about was just, the Google MCP, so I think it’s probably worth just, like, reaching out to them, since it’s been a number of days at this point.
245 00:26:53.420 ⇒ 00:26:58.210 Pranav Narahari: I think that…
246 00:26:58.380 ⇒ 00:27:06.599 zacfromson: I’ll reach out to them. I was told… when we were working with the other dev agency, they did say, like, to get this approved could take two to three weeks.
247 00:27:06.600 ⇒ 00:27:07.230 Pranav Narahari: Oh, God.
248 00:27:07.230 ⇒ 00:27:20.859 zacfromson: which I have no knowledge on, it’s just something they had said, whether it’s true or not. I don’t know, but I’ll reach out to them. I think it’s been about 2, so it is kind of getting up there. So let me… let me reach out to them. I’m the one who filled it out, so I’ll add that to my, my to-do list here.
249 00:27:22.200 ⇒ 00:27:24.900 Pranav Narahari: Okay, cool.
250 00:27:24.920 ⇒ 00:27:25.600 Samuel Roberts: Oh, go ahead.
251 00:27:26.320 ⇒ 00:27:32.190 Pranav Narahari: I was just gonna say, like, in terms of reports, like, that data is also useful for reports.
252 00:27:32.390 ⇒ 00:27:45.930 Pranav Narahari: So, we’ll just have to make sure we’re pulling in the right data once we can finally make that connection. And so, that’ll be, you know, probably, like, half day, a day’s work, once we kind of establish that MCP connection with Google.
253 00:27:48.250 ⇒ 00:27:48.820 zacfromson: Okay.
254 00:27:48.950 ⇒ 00:28:01.259 zacfromson: Cool. And then I think, just to make sure I’m clear on the whole convo, Bobby, too, I know we were talking yesterday, like, being able to get the knowledge base and, like, being able to upload our own prompts and Claude skills, like, probably be, like, the end of this, or do we gonna hold on that? What are your…
255 00:28:01.450 ⇒ 00:28:04.549 Bobby Palmieri: I think… I think, Sam, we probably need a way to get…
256 00:28:04.850 ⇒ 00:28:13.009 Bobby Palmieri: Claude’s skills in… on our end, just as a process thing. Like, what you’re uploading on the back end.
257 00:28:13.590 ⇒ 00:28:25.129 Bobby Palmieri: works great, but I think for us just to be able to paste that skill ID in, because I think that’s the path that we see moving forward. So, like, if I go…
258 00:28:25.560 ⇒ 00:28:27.570 Bobby Palmieri: Let me show my screen.
259 00:28:33.090 ⇒ 00:28:34.610 Samuel Roberts: Yeah, I think,
260 00:28:34.940 ⇒ 00:28:39.729 Samuel Roberts: my plan today was to try to take a look at that API and make sure we can fetch them, and .
261 00:28:39.990 ⇒ 00:28:47.069 Bobby Palmieri: I’m not even worried… I mean, you can spend a little bit of time on that, but honestly, just pasting… pasting in the skill ID into, like, some UI.
262 00:28:47.070 ⇒ 00:29:02.390 Samuel Roberts: Yeah, well, yeah, this is the thing, like, either way, we’ll need, like, some UI, so you can, like, turn them on and off if you want, or something, and we’re gonna need to change the database to make sure we store those IDs, and so those have to happen either way, and so the question is, like, can I pre-populate that for you, or not, you know?
263 00:29:03.030 ⇒ 00:29:10.659 Samuel Roberts: But yeah, the database to store the IDs, be able to, you know, maybe turn them off so that, like.
264 00:29:10.890 ⇒ 00:29:25.000 Samuel Roberts: they’re in there, but maybe Stitch doesn’t see them if you don’t want them to, so you can, like, play around with other ones on Claude or something, however that works. I’ll figure that out. And then, yeah, if we can fetch them and just have that list populated, great. If not, I’ll make it easy to, like, paste them right in, so…
265 00:29:25.330 ⇒ 00:29:26.010 Bobby Palmieri: Awesome.
266 00:29:26.610 ⇒ 00:29:40.029 Bobby Palmieri: Yeah, because I think it’s probably, like, again, I don’t know how familiar you guys are with Cloud Skills, but it’s really just, like, a markdown, long prompt with some examples and some output. So, like, you know, instead of doing a,
267 00:29:40.420 ⇒ 00:29:44.060 Bobby Palmieri: No, this did populate, I don’t… it looked like it was broken, so I…
268 00:29:44.460 ⇒ 00:29:46.679 Bobby Palmieri: Stop sharing my screen, but like…
269 00:29:46.990 ⇒ 00:29:51.780 Bobby Palmieri: you know, I… oh, this is the quiz flow analysis.
270 00:29:52.000 ⇒ 00:30:01.139 Bobby Palmieri: Oh, here’s the weekly meta report. So, like, this is, you know, something that I built in Cloud Skills of, like, the format that I want to use.
271 00:30:02.040 ⇒ 00:30:12.049 Bobby Palmieri: And it’s like, you know, it’s got the… the breakdown, it’s got the key insights, and, like, summary. So it’s really nice, and I think it’s probably the best way for us to spin up
272 00:30:12.330 ⇒ 00:30:17.159 Bobby Palmieri: quote-unquote agents or, like, tasks, because, like, Zach…
273 00:30:17.450 ⇒ 00:30:34.090 Bobby Palmieri: who’s closer to email than I am, like, he spent an hour building out, like, a Newton… or a email calendar claud still, of, like, what he wants to see, and then we can just use that on top of the other tool that we built. I could do a meta-report, I could do…
274 00:30:34.220 ⇒ 00:30:41.830 Bobby Palmieri: a Google report, we could do a monthly creative report, right? So, we can spend time in Claude’s skills building those out.
275 00:30:41.850 ⇒ 00:30:53.219 Bobby Palmieri: But Zach was running into, like, an issue last night where, like, he wasn’t able to necessarily test it. Yeah. So we’d need to send it to you, you’d add it, Zach would test it, go back with the feedback, so…
276 00:30:53.220 ⇒ 00:31:01.899 Bobby Palmieri: just the ability to be able to kind of rip those quickly. And then the third component of that is I sent a screenshot a little while ago of just, like.
277 00:31:01.990 ⇒ 00:31:06.580 Bobby Palmieri: how we want to maybe do the UI of, like, the prompt library.
278 00:31:07.250 ⇒ 00:31:21.120 Bobby Palmieri: So that those can be attached to, like, a Claude skill, right? So, like, when the team’s onboarded, they have the weekly report agent that they can just click, and it runs, or it schedules, or whatever, in that regard, too.
279 00:31:21.120 ⇒ 00:31:24.670 zacfromson: Yeah, and did you talk about the knowledge base as well? I don’t know if that’s been…
280 00:31:24.930 ⇒ 00:31:26.930 zacfromson: Haven’t got there. Okay.
281 00:31:26.930 ⇒ 00:31:30.269 Bobby Palmieri: What are you guys’ thoughts on, like, how complex, like…
282 00:31:31.060 ⇒ 00:31:37.650 Bobby Palmieri: adding a knowledge basis for, like, just another kind of tool call, so I can share my screen.
283 00:31:37.650 ⇒ 00:31:41.500 Pranav Narahari: Yeah, so I have some experience with knowledge bases, and…
284 00:31:41.700 ⇒ 00:31:53.390 Pranav Narahari: I’m wondering how you guys want it integrated, like, do you guys have a Google Drive and you want direct integration that way? Do you guys have it in some other place, or is it more like copy and pasting text?
285 00:31:53.680 ⇒ 00:31:56.299 Pranav Narahari: Like, how would you guys see that working?
286 00:31:56.460 ⇒ 00:31:58.230 Bobby Palmieri: This works really well for us.
287 00:31:58.720 ⇒ 00:32:04.129 Bobby Palmieri: This is just, like, a cloud project that we have on, like, a per-client basis. Okay.
288 00:32:04.350 ⇒ 00:32:12.290 Bobby Palmieri: And, like, essentially, I’m not entirely sure all the things that we upload here, and I think we probably wouldn’t upload all of these.
289 00:32:12.450 ⇒ 00:32:25.360 Bobby Palmieri: moving forward, but, like, you know, we have, like, their review articles, we have, like, a CSV of all of their reviews, so, like, as it’s generating emails, it can pull, like, actual reviews in for copy and.
290 00:32:25.360 ⇒ 00:32:25.820 Samuel Roberts: Sure.
291 00:32:25.820 ⇒ 00:32:26.840 Bobby Palmieri: as well.
292 00:32:27.290 ⇒ 00:32:27.890 Samuel Roberts: Okay.
293 00:32:30.170 ⇒ 00:32:42.879 Pranav Narahari: Yeah, Sam, I almost feel like just having another tab for knowledge bases where they can group them and, you know, yeah, email copy versus, you know, whatever other context, and then it’s almost…
294 00:32:42.970 ⇒ 00:32:50.929 Pranav Narahari: Like, you can turn on a slider, or you can turn on a switch for a specific chat window.
295 00:32:52.020 ⇒ 00:32:54.140 Pranav Narahari: I could see that working well.
296 00:32:54.400 ⇒ 00:32:56.590 Samuel Roberts: Yeah, no, I see exactly what you’re saying there.
297 00:32:56.590 ⇒ 00:32:56.930 Pranav Narahari: Yeah.
298 00:32:58.890 ⇒ 00:32:59.950 Samuel Roberts: So, like.
299 00:33:00.250 ⇒ 00:33:05.930 Samuel Roberts: Just bear with me for a moment. So, like, that’s all… that’s all in Clawed in a project, right? You want something in Stitch.
300 00:33:06.460 ⇒ 00:33:07.870 Samuel Roberts: Where you upload…
301 00:33:08.160 ⇒ 00:33:15.809 Bobby Palmieri: We can’t call that. Zach and I can get you a little bit more context on, like, what we probably want to have in the knowledge base.
302 00:33:15.810 ⇒ 00:33:16.190 Samuel Roberts: Yeah.
303 00:33:16.860 ⇒ 00:33:23.060 zacfromson: Really just being able to upload, like, probably documents, where it’s like, here’s information about the brand that we want to add,
304 00:33:23.410 ⇒ 00:33:25.109 zacfromson: Things about, like, here’s, like, you know.
305 00:33:25.300 ⇒ 00:33:45.240 zacfromson: 50 emails we’ve already written for the brand, so we can, like, better understand the brand voice, just, like, things, like, knowledge that would be helpful on brands research we’ve done, things of that nature. Like, not saying it can’t, but if we have some precise stuff we want to feed it, it wouldn’t hurt just to, like, infuse some… basically just PDFs, Word docs.
306 00:33:45.240 ⇒ 00:33:50.450 Bobby Palmieri: The only thing that me and you probably need to think about More is…
307 00:33:50.730 ⇒ 00:33:58.580 Bobby Palmieri: what is a Claude skill versus what is, like, actually needed in the knowledge base? Because, like, I created the Newton…
308 00:33:59.350 ⇒ 00:34:01.440 Bobby Palmieri: Email copywriter skill.
309 00:34:01.890 ⇒ 00:34:02.460 Samuel Roberts: Like…
310 00:34:02.460 ⇒ 00:34:09.129 Bobby Palmieri: that’s where I uploaded, like, past emails, approved copy, right? And it, like, made its own tone doc.
311 00:34:09.790 ⇒ 00:34:16.620 Bobby Palmieri: So we should just chat about, like, what can be a skill, and, like, what do the tools need to call on, like, an ongoing basis?
312 00:34:16.620 ⇒ 00:34:16.940 zacfromson: Well.
313 00:34:17.150 ⇒ 00:34:19.920 Bobby Palmieri: So we can kind of map that out. I think if we…
314 00:34:19.920 ⇒ 00:34:23.090 Samuel Roberts: I think the Claude skills will probably get us 70% of the way there.
315 00:34:23.090 ⇒ 00:34:23.510 zacfromson: I agree.
316 00:34:23.510 ⇒ 00:34:27.169 Bobby Palmieri: Probably want, like, a few more things that it can pull in.
317 00:34:27.179 ⇒ 00:34:38.229 zacfromson: Yeah, and I think our main objective with all this is to try to, like… there’s gonna be more work we can continue to do on these agents, improve them over time, and add more. I think we want to kind of figure out, alright, like, what are the last couple things we want to sure up here?
318 00:34:38.809 ⇒ 00:34:49.439 zacfromson: So you guys are not distracted with that anymore. Can move on to forecasting the other projects we have, and get us, like, hey, this is good enough, we can roll this out, we have everything we need to really scale this with the team.
319 00:34:49.439 ⇒ 00:35:05.619 zacfromson: get a bunch of feedback, and then we can loop back in 30 days or 60 days once we get a lot of these other builds done, so we don’t keep, like, slowing ourselves down with, like, trying to add and add and add, because we can continue to feature build this all day, but it’s, like, what are the final things that I think are holding us back for, like.
320 00:35:06.019 ⇒ 00:35:08.849 zacfromson: A really good usability to let you guys just…
321 00:35:09.339 ⇒ 00:35:13.119 zacfromson: Pit pause on that, not talk about that, and… Move forward.
322 00:35:13.440 ⇒ 00:35:19.089 Pranav Narahari: Gotcha. So those two things, basically, it’s, having the Claude skills and the knowledge bases, that would kind of.
323 00:35:19.090 ⇒ 00:35:30.399 zacfromson: Yeah, or even… I don’t know how long the knowledge base is gonna take. Again, if it’s something that’s not super complicated, yes. I think just being able to add our own prompts and our own Cloud skills would be, like, the last big thing, in my opinion.
324 00:35:30.400 ⇒ 00:35:30.910 Bobby Palmieri: Goodbye.
325 00:35:31.140 ⇒ 00:35:31.520 Pranav Narahari: Okay.
326 00:35:31.520 ⇒ 00:35:45.929 Bobby Palmieri: My take on that is the image generation tool is good to go. MCP is working, like, I think we could probably roll this out next week, Zach, and, like, I think the next things that we need to be able to add are…
327 00:35:46.160 ⇒ 00:35:49.279 Bobby Palmieri: the knowledge… er, sorry, the Claude skills, like.
328 00:35:49.630 ⇒ 00:35:58.560 Bobby Palmieri: let’s just get that done. I think knowledge base can probably come later, because I think there’s a good amount of things in the knowledge base that can be done as a cloud skill, and probably…
329 00:35:58.560 ⇒ 00:36:06.870 zacfromson: upload, like, here’s all the emails we’ve ever written for this brand into the Claude skill, and then that could encapsulate the knowledge base. So I’d say, like, let’s…
330 00:36:07.340 ⇒ 00:36:15.970 zacfromson: basically get… being able to add our own prompts and Claude skills, and being able to connect Google, and I think we’re probably there, Bobby.
331 00:36:15.970 ⇒ 00:36:16.310 Samuel Roberts: Yeah.
332 00:36:16.310 ⇒ 00:36:19.229 Bobby Palmieri: I mean, we’ll roll this out before we get Google.
333 00:36:19.230 ⇒ 00:36:24.989 zacfromson: Correct. I think we can roll this out even without clod skills. Like, I’m already using it for some polls, but I think, like.
334 00:36:25.430 ⇒ 00:36:30.790 zacfromson: for Brain Forge, like, do those two last things, and then, like, focus on forward features.
335 00:36:31.010 ⇒ 00:36:41.860 Samuel Roberts: Yeah. Yeah, skills shouldn’t be crazy, and I think you guys are on the right track there. Like, if it’s something that’s gonna be repeating and not, like, the underlying data is not gonna change a lot, the cloud skill is a better way to go than…
336 00:36:42.090 ⇒ 00:36:44.350 Samuel Roberts: Focusing on the knowledge base, because, like.
337 00:36:45.010 ⇒ 00:36:56.519 Samuel Roberts: Yeah, that’s gonna run every time, and, you know, these things are non-deterministic, so things will change, but if you can, like, solidify it into a skill that has, like, examples that are generated from previous things, and then that’s what you run, I think that’s perfect.
338 00:36:56.930 ⇒ 00:37:03.029 Bobby Palmieri: Yeah, because I think, like, for the most part, the knowledge is static, Zach, so I think we’d actually be better off…
339 00:37:03.290 ⇒ 00:37:05.250 Bobby Palmieri: Doing it in a Claude scale, right?
340 00:37:05.250 ⇒ 00:37:10.790 zacfromson: I agree. I think we can… yeah, let’s ignore the knowledge for now. Let’s revisit that, you know? Yeah.
341 00:37:10.790 ⇒ 00:37:11.309 Samuel Roberts: Yeah, I mean…
342 00:37:11.310 ⇒ 00:37:23.679 zacfromson: But I think for now, like, you know, if I make a, a copywriting agent for X brand, I’m gonna upload all of the past copy, all the brief guidelines into that skill.
343 00:37:23.680 ⇒ 00:37:30.730 Bobby Palmieri: And it’s gonna be encapsulated there anyways. Zach, if that changes over time, you can just re-prompt the skill and re-upload it?
344 00:37:30.730 ⇒ 00:37:40.669 zacfromson: Correct. Yep. So let’s just build our own prompts and add our own skills, right? So I think if we want to have, like, a general prompt, Bobby, like, get me a report for the last 7 days, just give the team ideas.
345 00:37:41.680 ⇒ 00:37:49.750 zacfromson: like, text-based prompts? Or do you want to not even do that, just get cloud skills that we can manage and add on a per… basically global and per-account basis, and then move on?
346 00:37:49.750 ⇒ 00:37:59.049 Bobby Palmieri: Yeah, I just tagged you and Sam. I’ll tag you in this as well. On Slack is, like, how I’m thinking about it.
347 00:37:59.830 ⇒ 00:38:08.059 Bobby Palmieri: You know, in terms of, like, the prompt library, which would really be the agent library, which would be, like, you know, the weekly performance report is a Claude skill.
348 00:38:08.630 ⇒ 00:38:11.970 Bobby Palmieri: You can either run it now or schedule it to send to Slack.
349 00:38:13.210 ⇒ 00:38:21.609 Bobby Palmieri: you know, in that regard. So, Sam, we can figure out what step one is and what the final end state is, but that’s how I’m thinking about, like, where we want to go with this, is…
350 00:38:21.760 ⇒ 00:38:40.869 Bobby Palmieri: Zach and I are really, like, the team can have free use of Stitch, but Zach and I are really gonna build out, like, the prompts that we think are impactful, and then the team should use those. So, like, I have this meta weekly report, Claude Skill, that’s a quote-unquote, and we’ll frame it as an agent to the team.
351 00:38:41.720 ⇒ 00:38:50.289 Bobby Palmieri: they can, you know, see that under the prompt library, or we’ll call it the agent library, and they can just run that, and it’ll just be the prompt with the Claude skill.
352 00:38:50.810 ⇒ 00:38:52.479 Samuel Roberts: And they’ll get their output.
353 00:38:53.650 ⇒ 00:38:54.910 Samuel Roberts: Yeah, I think that makes sense.
354 00:38:56.970 ⇒ 00:38:59.939 Samuel Roberts: Okay. Yeah, the skills, like I said, I should be able to hopefully
355 00:39:00.310 ⇒ 00:39:04.610 Samuel Roberts: Knock that out in one of those two forms pretty quickly, and get that in.
356 00:39:05.250 ⇒ 00:39:05.810 Bobby Palmieri: Sweet.
357 00:39:07.190 ⇒ 00:39:07.770 Pranav Narahari: Cool.
358 00:39:09.440 ⇒ 00:39:20.139 Pranav Narahari: Yeah, I think we’re all basically aligned on the nana banana stuff. I think Casey, yeah, and Bobby, you said, like, that’s in a good spot. Casey, is there anything you want to mention there?
359 00:39:20.780 ⇒ 00:39:21.840 Pranav Narahari: I think we’re good, right?
360 00:39:22.740 ⇒ 00:39:25.820 Casie Aviles: Yeah, and if there’s, like, any updates or…
361 00:39:25.980 ⇒ 00:39:36.129 Casie Aviles: Any ideas that we want, it would be just good to, like, create a backlog for that. But yeah, otherwise, yeah, that’s all I have for the image gene.
362 00:39:36.430 ⇒ 00:39:41.890 Bobby Palmieri: Yeah, the… the… Next, there’s two features there,
363 00:39:43.600 ⇒ 00:39:47.440 Bobby Palmieri: That we’re gonna wanna add is… the photoshoot.
364 00:39:49.230 ⇒ 00:39:52.170 Bobby Palmieri: And then probably a reference gallery.
365 00:39:52.600 ⇒ 00:39:53.350 Samuel Roberts: Hmm.
366 00:39:55.290 ⇒ 00:39:59.689 Bobby Palmieri: Like, I don’t… like, we already have the reference image that is working well.
367 00:39:59.810 ⇒ 00:40:08.850 Bobby Palmieri: So it’s just a Pinterest board, if you will, where you can just select that to use as a reference image.
368 00:40:10.420 ⇒ 00:40:13.509 Bobby Palmieri: So I can get that written up for you.
369 00:40:15.260 ⇒ 00:40:16.060 Casie Aviles: Okay.
370 00:40:18.110 ⇒ 00:40:19.110 Pranav Narahari: Perfect, yup.
371 00:40:19.380 ⇒ 00:40:20.650 Pranav Narahari: Little,
372 00:40:21.110 ⇒ 00:40:27.120 Pranav Narahari: We’ll look for that in Slack, and then, yeah, like Casey said, we’ll add those things to the backlog as well.
373 00:40:27.870 ⇒ 00:40:31.329 Pranav Narahari: Cool, that was, the agenda for today.
374 00:40:32.050 ⇒ 00:40:34.800 Pranav Narahari: I think we’re good, you know, we’re 20 minutes early, but…
375 00:40:35.050 ⇒ 00:40:39.069 Pranav Narahari: you know, the main thing was, yeah, production’s live. I think once you guys can, like.
376 00:40:39.260 ⇒ 00:40:48.100 Pranav Narahari: dive into there. I’m sure you guys will have some feedback for us. We’ve already had some notes that we’re gonna already be pushing out later today. But yeah.
377 00:40:48.880 ⇒ 00:40:53.030 Pranav Narahari: yeah, start poking around there and let us know, what you guys come up with. Cool.
378 00:40:53.030 ⇒ 00:40:56.690 Bobby Palmieri: So I think the big things, Claude Scale upload.
379 00:40:57.440 ⇒ 00:41:02.300 Bobby Palmieri: Let us know how close or how far we are on the two features that we vibe-coded.
380 00:41:03.210 ⇒ 00:41:05.380 Bobby Palmieri: And then…
381 00:41:06.000 ⇒ 00:41:15.589 Bobby Palmieri: I think, yeah, we’re in a good spot. Pranav, could you, just before end of day, let us know kind of where we stand? I know we’re just getting started on the forecasting UI.
382 00:41:16.140 ⇒ 00:41:17.889 Bobby Palmieri: We’re kind of, like, where we’re at with that.
383 00:41:18.940 ⇒ 00:41:23.989 Pranav Narahari: Yeah, definitely. By end of day, we’ll have, like, updates on that, for sure.
384 00:41:25.160 ⇒ 00:41:25.980 Bobby Palmieri: Sweet.
385 00:41:26.900 ⇒ 00:41:27.610 Pranav Narahari: Perfect.
386 00:41:28.480 ⇒ 00:41:29.650 Pranav Narahari: Alright, guys.
387 00:41:29.890 ⇒ 00:41:31.650 Bobby Palmieri: Have a good rest of your day.
388 00:41:32.000 ⇒ 00:41:33.040 Bobby Palmieri: Go Patriots!
389 00:41:33.210 ⇒ 00:41:33.780 Pranav Narahari: Yeah, go ahead.
390 00:41:33.780 ⇒ 00:41:35.160 Bobby Palmieri: Patriots.
391 00:41:35.480 ⇒ 00:41:36.410 Bobby Palmieri: Okay, got too much.
392 00:41:37.100 ⇒ 00:41:38.029 Pranav Narahari: See you guys.
393 00:41:38.030 ⇒ 00:41:38.420 Samuel Roberts: Thank you guys.
394 00:41:38.420 ⇒ 00:41:39.050 zacfromson: Yes.