Meeting Title: Brainforge Platform Development Sync Date: 2026-03-03 Meeting participants: Samuel Roberts, Uttam Kumaran
WEBVTT
1 00:07:58.560 ⇒ 00:07:59.790 Uttam Kumaran: Hello, sir.
2 00:08:00.140 ⇒ 00:08:02.120 Samuel Roberts: Hey, I was just about to ping you.
3 00:08:02.120 ⇒ 00:08:04.340 Uttam Kumaran: Sorry, I just got this interview, just finished.
4 00:08:04.340 ⇒ 00:08:11.800 Samuel Roberts: Oh, I forgot, yeah, I did forget about… okay, I did see that after I said, yeah, my bad. I knew that was… should’ve remembered. I just got off an interview at 4, too, so I’m.
5 00:08:11.800 ⇒ 00:08:12.710 Uttam Kumaran: How was it?
6 00:08:13.440 ⇒ 00:08:15.150 Samuel Roberts: It was good! It was, I didn’t…
7 00:08:15.400 ⇒ 00:08:25.480 Samuel Roberts: I had two today, and I was not aware of them until I got on this morning, so I was a little bit like, but, they were good. One of them was better than the other, I think, in past.
8 00:08:26.900 ⇒ 00:08:32.839 Uttam Kumaran: Okay. I mean, I’m actually not expecting them all to go well. So…
9 00:08:32.840 ⇒ 00:08:35.939 Samuel Roberts: Well, yeah, go well, yeah, I mean, nothing’s been atrocious.
10 00:08:35.940 ⇒ 00:08:36.459 Uttam Kumaran: Yeah, yeah, yeah.
11 00:08:36.460 ⇒ 00:08:40.680 Samuel Roberts: that way. But yeah, one of them, he just didn’t feel quite right. Okay.
12 00:08:40.890 ⇒ 00:08:45.840 Samuel Roberts: the other one felt a bit better, and… I mean, this first one is… is…
13 00:08:46.010 ⇒ 00:08:50.609 Samuel Roberts: not the technical one, so I’m still curious about some of that, you know, but… Okay. Yeah.
14 00:08:51.560 ⇒ 00:08:55.129 Uttam Kumaran: Cool, dude, nice. Yeah, overall, you know… How’s time off?
15 00:08:55.520 ⇒ 00:08:58.709 Samuel Roberts: Oh, it was good. It was a busy, busy weekend.
16 00:08:59.460 ⇒ 00:09:01.919 Samuel Roberts: We flew to Arizona for a wedding.
17 00:09:01.920 ⇒ 00:09:03.350 Uttam Kumaran: Yeah, with a kid, right?
18 00:09:03.350 ⇒ 00:09:18.230 Samuel Roberts: Yeah, and so we had been in Florida the week before, and that was an okay flight. This one was a connecting flight, so I was like, okay, great, just, like, twice the potential for bad. But he did great. All my flights were good, he was a hit at the wedding.
19 00:09:19.220 ⇒ 00:09:24.290 Samuel Roberts: My friends loved him, you know, everything was… he’s a good kid most of the time, so it’s hard to…
20 00:09:24.570 ⇒ 00:09:24.990 Uttam Kumaran: Nice.
21 00:09:24.990 ⇒ 00:09:26.019 Samuel Roberts: Gotta mess that up.
22 00:09:26.280 ⇒ 00:09:26.810 Samuel Roberts: But…
23 00:09:26.810 ⇒ 00:09:27.550 Uttam Kumaran: Nice.
24 00:09:27.550 ⇒ 00:09:28.260 Samuel Roberts: Yeah.
25 00:09:29.420 ⇒ 00:09:42.410 Samuel Roberts: But yeah, busy, busy flying back. I mean, going out there is… there’s still, like, 11 hours of travel, but you get there and you have the day, whereas coming back, you leave and, like, lose the whole day. It’s just…
26 00:09:42.990 ⇒ 00:09:45.370 Samuel Roberts: So I got home last night and was like, oh, man.
27 00:09:47.160 ⇒ 00:09:51.200 Uttam Kumaran: Yeah, I… my weekend was a lot of this, which was great.
28 00:09:51.200 ⇒ 00:09:51.810 Samuel Roberts: Yeah?
29 00:09:51.810 ⇒ 00:09:52.649 Uttam Kumaran: Oh, good, good.
30 00:09:52.760 ⇒ 00:09:55.069 Uttam Kumaran: I feel like I hit, like…
31 00:09:55.390 ⇒ 00:09:57.679 Uttam Kumaran: I’m, like, in warp speed right now.
32 00:09:57.680 ⇒ 00:09:58.869 Samuel Roberts: Yeah, yeah. Interesting.
33 00:09:59.060 ⇒ 00:10:01.110 Uttam Kumaran: Like, I think… I think I’ve reached…
34 00:10:01.480 ⇒ 00:10:05.399 Uttam Kumaran: I can’t work on more things right now, like…
35 00:10:05.400 ⇒ 00:10:06.010 Samuel Roberts: Nice.
36 00:10:06.490 ⇒ 00:10:09.069 Uttam Kumaran: Meaning, I’m not writing a lick of code myself.
37 00:10:09.070 ⇒ 00:10:09.570 Samuel Roberts: Yeah.
38 00:10:10.220 ⇒ 00:10:16.849 Samuel Roberts: I’m mainly troubleshooting and keys, which I can tell you are, too. Yeah, that seems… I mean, especially, like, you know.
39 00:10:17.880 ⇒ 00:10:24.140 Samuel Roberts: I was saying this earlier, like, you know, certain things don’t live in the repo, and you can’t just cursor your way out of them, but… Yeah, so, like…
40 00:10:24.140 ⇒ 00:10:29.250 Uttam Kumaran: I mean, I think we need to think of something for the MV keys, because I’ve been maintaining, basically.
41 00:10:29.790 ⇒ 00:10:35.090 Uttam Kumaran: Cursor Cloud, My End Local, and the 1Pass.
42 00:10:35.310 ⇒ 00:10:43.460 Uttam Kumaran: But, like, as you can see, like, I think you asked, is there a Clockify API key? The thing with the Clockify API key, though, is, like, it’s on an individual basis.
43 00:10:43.460 ⇒ 00:10:45.890 Samuel Roberts: I was one… that’s kind of… I saw, it was like…
44 00:10:45.890 ⇒ 00:10:53.410 Uttam Kumaran: I’ll send you mine, but, like, I think I’m gonna… I have to… I basically just wanted to get this working before I, like… I need to do another visual pass.
45 00:10:53.610 ⇒ 00:10:54.130 Samuel Roberts: Okay.
46 00:10:54.130 ⇒ 00:10:59.009 Uttam Kumaran: add it to, like, add helpful tool tips, but I can give you one to test, but, like, yeah, it’s… it’s on a.
47 00:10:59.010 ⇒ 00:11:04.569 Samuel Roberts: I wasn’t sure if there was an OAuth option for that, too, or not. I didn’t know how Clockify did it yet, I hadn’t dug in.
48 00:11:04.570 ⇒ 00:11:07.949 Uttam Kumaran: I don’t think we have SSO set up, so it’s all API key.
49 00:11:08.850 ⇒ 00:11:09.460 Samuel Roberts: Oh, okay.
50 00:11:09.460 ⇒ 00:11:13.199 Uttam Kumaran: It’s a one-time step, like, if people just can go grab it real quick, you know?
51 00:11:13.200 ⇒ 00:11:15.320 Samuel Roberts: Yeah, yeah, okay.
52 00:11:16.000 ⇒ 00:11:20.840 Samuel Roberts: Alright, I won’t worry too much about that then until you’re,
53 00:11:21.540 ⇒ 00:11:25.079 Samuel Roberts: We can figure that out once you finish the, like, visual stuff, I guess.
54 00:11:25.370 ⇒ 00:11:33.009 Samuel Roberts: I’m trying to see… I’m just in Clockify now. Oh, there’s Manager PI keys, yeah, okay. All right, yeah, maybe I’ll throw one down and see what I can play with it a little bit.
55 00:11:34.140 ⇒ 00:11:34.790 Uttam Kumaran: Nice.
56 00:11:35.200 ⇒ 00:11:35.850 Samuel Roberts: Yeah.
57 00:11:37.480 ⇒ 00:11:42.580 Samuel Roberts: There was another… that triggered another thought, and I just got distracted by Clockify and don’t remember what it was now.
58 00:11:44.550 ⇒ 00:11:53.390 Samuel Roberts: Something… Oh, yeah, I was doing some reading about how best to manage Environment variables in general.
59 00:11:53.410 ⇒ 00:11:54.650 Uttam Kumaran: Okay.
60 00:11:54.980 ⇒ 00:12:01.249 Samuel Roberts: Some… ugh, this was some post on Bookface I saw, but it was also, like, on my phone over the weekend, so I don’t think I have it up here.
61 00:12:01.840 ⇒ 00:12:08.009 Samuel Roberts: But people were literally discussing this exact topic, because everyone’s dealing with every… you know what I mean?
62 00:12:09.150 ⇒ 00:12:11.919 Samuel Roberts: Especially coding agents, let me see if I can find it real quick.
63 00:12:15.070 ⇒ 00:12:23.069 Samuel Roberts: I forget what people said… But I was gonna… I meant to do some digging into, like, 1Password…
64 00:12:23.780 ⇒ 00:12:27.470 Samuel Roberts: in general, like the CLI, I haven’t played with that much myself.
65 00:12:29.730 ⇒ 00:12:35.129 Samuel Roberts: Because people were using other… other tools for that, but I thought, you know, if we’re already using it.
66 00:12:35.130 ⇒ 00:12:36.850 Uttam Kumaran: Well, I don’t know, like, yeah, I mean…
67 00:12:37.340 ⇒ 00:12:39.940 Uttam Kumaran: I just don’t know how to, like, keep it in sync.
68 00:12:41.010 ⇒ 00:12:45.850 Uttam Kumaran: like, and I don’t… I mean, one pass, like, yes, you could save it there, but, like, I’m wondering, like, should we…
69 00:12:47.130 ⇒ 00:12:50.409 Uttam Kumaran: Do we actually save it to the repo as, like, the central spot?
70 00:12:51.720 ⇒ 00:12:52.790 Uttam Kumaran: I don’t know.
71 00:12:53.280 ⇒ 00:12:54.300 Samuel Roberts: Yeah, I think…
72 00:12:55.050 ⇒ 00:13:00.799 Samuel Roberts: I don’t know, there’s definitely other tools out there that we could look at, especially if we’re gonna be, you know, everyone’s gonna be running stuff, and…
73 00:13:00.900 ⇒ 00:13:06.079 Samuel Roberts: We’re gonna have lots of environments out there, locally, in the cloud, and railway, and everything.
74 00:13:06.080 ⇒ 00:13:07.180 Uttam Kumaran: Okay.
75 00:13:07.180 ⇒ 00:13:09.890 Samuel Roberts: But I… I think there’s… there’s some…
76 00:13:10.710 ⇒ 00:13:16.259 Samuel Roberts: you know, there’s secret management tools, I just… I haven’t used many of them, because I haven’t worked on such a, you know…
77 00:13:16.510 ⇒ 00:13:16.830 Uttam Kumaran: Yeah.
78 00:13:16.830 ⇒ 00:13:22.200 Samuel Roberts: so many different people and different tools yet, but, I can do some digging on that and see what the best
79 00:13:22.910 ⇒ 00:13:27.029 Samuel Roberts: would be… I mean, I’d set up something on Heroku, er, with a Heroku…
80 00:13:27.260 ⇒ 00:13:37.379 Samuel Roberts: CLI to pull down environment variables as just, like, a script for people to use. So we might be able to do something like that with Railway. I don’t think Railway necessarily has…
81 00:13:38.370 ⇒ 00:13:43.640 Samuel Roberts: Like, a built-in, you know, secret manager, but because we’re storing them there anyway, we could probably still pull them down.
82 00:13:43.740 ⇒ 00:13:48.100 Samuel Roberts: The only issue becomes, like, some of the custom ones that are up there for, like, the internal.
83 00:13:48.230 ⇒ 00:13:49.410 Samuel Roberts: networking.
84 00:13:49.680 ⇒ 00:13:54.380 Samuel Roberts: But… But I can… I can put some thought to that, because I think that’s definitely a…
85 00:13:54.800 ⇒ 00:13:58.530 Samuel Roberts: an ongoing thing. There was an interesting article about,
86 00:13:58.650 ⇒ 00:14:09.930 Samuel Roberts: about some of this that I read from 1Password on, like, Claudebot stuff, where they were, like, talking about how to lock things down and stuff like that, so I feel like they’re thinking about it, at least, so maybe it’ll be something soon, but we can probably hack our way around it for now.
87 00:14:10.530 ⇒ 00:14:11.510 Uttam Kumaran: Yeah, okay.
88 00:14:13.810 ⇒ 00:14:14.380 Samuel Roberts: Cool.
89 00:14:17.840 ⇒ 00:14:28.999 Samuel Roberts: Oh, yeah, I have not had a chance to, like, get my thoughts down in terms of testing, but one thing I was thinking about, this is just top of mind, like, if we chat tomorrow with Pranav, like.
90 00:14:29.270 ⇒ 00:14:34.979 Samuel Roberts: Listing out the… I mean, I could probably do this, but I also want, like, your input on, like.
91 00:14:35.170 ⇒ 00:14:37.819 Samuel Roberts: The, like.
92 00:14:37.960 ⇒ 00:14:43.459 Samuel Roberts: basic things we want to make sure are, like, the first things we test for the platform, because, like I said, I’m more willing to let
93 00:14:43.940 ⇒ 00:14:49.889 Samuel Roberts: other PRs through for new features that might not be fully tested, as long as we don’t break…
94 00:14:50.170 ⇒ 00:14:50.980 Samuel Roberts: You know.
95 00:14:51.130 ⇒ 00:14:54.429 Samuel Roberts: The basic things, like meeting search and stuff, things like that.
96 00:14:54.590 ⇒ 00:14:54.980 Uttam Kumaran: Yeah.
97 00:14:54.980 ⇒ 00:15:04.529 Samuel Roberts: I wasn’t sure if there were other things that you were like, this is definitely one we need to make sure it doesn’t break, this is definitely, you know, at least in an order. Obviously, everything is eventually gonna get there, ideally, but.
98 00:15:04.850 ⇒ 00:15:07.979 Uttam Kumaran: I mean, really, the biggest thing for me is that, like, you can log in.
99 00:15:07.990 ⇒ 00:15:08.710 Samuel Roberts: Okay.
100 00:15:09.510 ⇒ 00:15:13.660 Uttam Kumaran: It’s like, login, yeah, it’s like…
101 00:15:13.850 ⇒ 00:15:16.690 Uttam Kumaran: Basically, being able to click into a meeting.
102 00:15:17.080 ⇒ 00:15:17.640 Samuel Roberts: Okay.
103 00:15:17.640 ⇒ 00:15:19.959 Uttam Kumaran: and meeting search, I think, is fine for now.
104 00:15:20.130 ⇒ 00:15:28.119 Samuel Roberts: Okay, that’s fine. Yeah, I wasn’t… that was sort of the initial thing, I was like, this seems like the most used thing, so I didn’t want to assume that there wasn’t anything else, but…
105 00:15:28.430 ⇒ 00:15:30.629 Samuel Roberts: I can… okay, cool.
106 00:15:31.730 ⇒ 00:15:36.749 Samuel Roberts: So that should be… I mean, I think the easiest way to do that is probably just, like, mock out some, like, superbase endpoints and…
107 00:15:36.910 ⇒ 00:15:38.930 Samuel Roberts: Just make sure everything loads.
108 00:15:39.450 ⇒ 00:15:43.000 Samuel Roberts: Even if we don’t do, like, a full, like, playwright end-to-end test.
109 00:15:43.310 ⇒ 00:15:43.650 Uttam Kumaran: Okay.
110 00:15:43.650 ⇒ 00:15:48.189 Samuel Roberts: Just make sure… make sure no one, like, messed with anything that’s gonna break that. It’s kind of my thought.
111 00:15:48.240 ⇒ 00:15:49.260 Uttam Kumaran: Okay.
112 00:15:49.420 ⇒ 00:15:50.610 Samuel Roberts: I think that’ll be good.
113 00:15:51.150 ⇒ 00:15:55.360 Samuel Roberts: Yeah.
114 00:15:56.660 ⇒ 00:16:05.590 Samuel Roberts: Yeah, how are you doing? How… like, you seem to be, like, digging into the stuff, like, are you feeling, like, better about, like, getting some of this stuff done? Like, getting, like, less in the… I don’t know.
115 00:16:05.590 ⇒ 00:16:11.389 Uttam Kumaran: Yeah, I… well, I mean, I love this. I feel like I’m just, like, a kid, like, I feel like.
116 00:16:11.390 ⇒ 00:16:11.890 Samuel Roberts: Yeah.
117 00:16:11.890 ⇒ 00:16:13.040 Uttam Kumaran: Back to, like…
118 00:16:13.540 ⇒ 00:16:23.450 Uttam Kumaran: When I started my career, but this is great. I think, for me, what I’m trying to just do is, like, one, of course, I’m trying to ship the backlog of things that we’ve wanted for the
119 00:16:24.540 ⇒ 00:16:29.290 Uttam Kumaran: And I think I finally am at a place where I can… where I can do that solo. Cool.
120 00:16:29.500 ⇒ 00:16:32.399 Uttam Kumaran: Which I feel really, really good about.
121 00:16:33.460 ⇒ 00:16:36.899 Uttam Kumaran: You know, so that’s, like, one thing. I think second thing is.
122 00:16:37.270 ⇒ 00:16:45.670 Uttam Kumaran: This monorepo, and making… giving… putting the transcripts in there has been a… A step function for everybody.
123 00:16:45.670 ⇒ 00:16:46.610 Samuel Roberts: Yeah.
124 00:16:46.610 ⇒ 00:16:57.540 Uttam Kumaran: like, I think the decision to do that, and the decision to, like, not wait for, like, something UI-based, like, just to give cursor to everybody, everybody is, like.
125 00:16:57.940 ⇒ 00:17:07.380 Uttam Kumaran: doing really, really, really well. And Brile also, you know, you’ll find it interesting, his background is not in… he’s not technical at all.
126 00:17:07.750 ⇒ 00:17:12.930 Uttam Kumaran: He has no formal technical background, and on his first day, I basically said.
127 00:17:13.040 ⇒ 00:17:17.800 Uttam Kumaran: the first meeting with me was, like, let’s… we’re gonna start your… your job here all through Cursor.
128 00:17:17.800 ⇒ 00:17:18.960 Samuel Roberts: Yeah.
129 00:17:19.109 ⇒ 00:17:22.619 Uttam Kumaran: And so, he… and within 3 days, he was able to…
130 00:17:22.769 ⇒ 00:17:28.039 Uttam Kumaran: basically write his own case study and SOW with no understanding of what our company does.
131 00:17:28.519 ⇒ 00:17:29.749 Samuel Roberts: That’s crazy. Like, yeah.
132 00:17:29.750 ⇒ 00:17:32.530 Uttam Kumaran: I just was like, hey, I need to do this, can you go, can you just, like.
133 00:17:32.750 ⇒ 00:17:37.139 Uttam Kumaran: take this ask and, like, use Kirscher and figure out, like, what it is someone’s asking. And he did it.
134 00:17:37.510 ⇒ 00:17:38.060 Samuel Roberts: I love that.
135 00:17:38.060 ⇒ 00:17:44.070 Uttam Kumaran: So that really, like, kind of was, like, a little frame-breaking for me, where I was like, okay, like…
136 00:17:45.430 ⇒ 00:17:48.590 Uttam Kumaran: we’re, like, we’re… we’re… I think, like.
137 00:17:48.990 ⇒ 00:17:56.960 Uttam Kumaran: really, we’re the ones in our own way. And so, like, and I think, like, this is even where it can happen in some ways. Like, I still think that…
138 00:17:57.460 ⇒ 00:18:06.370 Uttam Kumaran: I think… I’m sure, like, a lot of the engineers are just not pushing it to the limit, because we’re just used to, like, developing a certain way, right? Like, we’re not used to.
139 00:18:06.370 ⇒ 00:18:06.770 Samuel Roberts: developing.
140 00:18:06.770 ⇒ 00:18:12.420 Uttam Kumaran: Where it’s, like, assigning a… like, spending most of your time actually just writing the plan and creating the tickets.
141 00:18:12.540 ⇒ 00:18:14.350 Uttam Kumaran: And signing it off, you know?
142 00:18:14.350 ⇒ 00:18:29.830 Samuel Roberts: Totally. No, I… it’s interesting, because I find myself, like, developing in different ways sometimes, and I’m like, oh, I’m not… I’m getting too deep into it, and I have to step back, like… like, last week, I was helping Gabe with the linear ticket thing, and, like, debugging stuff, I’m just… I, like, I kept being like, oh, wait, let’s just feed this to Cursor and see what it does.
143 00:18:29.830 ⇒ 00:18:30.279 Uttam Kumaran: Yeah, yeah.
144 00:18:30.280 ⇒ 00:18:39.869 Samuel Roberts: So in the, like, debugging mode, my brain wasn’t there. Whereas, like, sometimes, like, on a couple side projects I was just hacking on, like, I haven’t looked at the code that got written, you know? I need to get…
145 00:18:40.990 ⇒ 00:18:46.929 Samuel Roberts: Closer to that, at least. Maybe not completely, but… Yeah, so that’s exactly where I’m like, okay, where does our time shift?
146 00:18:46.930 ⇒ 00:18:53.550 Uttam Kumaran: So I’m kind of, like, figuring that out. But we also still have a backlog of stuff that I… we just need to do for the company that I’m just trying to, like…
147 00:18:53.900 ⇒ 00:19:00.280 Uttam Kumaran: get into a good place, and so I’m breaking, kind of, as much… I’m trying to, like, see what it’s like for me to ship all this stuff.
148 00:19:00.560 ⇒ 00:19:01.120 Samuel Roberts: Yeah.
149 00:19:01.510 ⇒ 00:19:08.720 Uttam Kumaran: And what breaks, and so that’s why it’s like, okay, like, if you look, I’m shipping, like, 10, 15, 20 PRs a day now.
150 00:19:09.460 ⇒ 00:19:15.230 Uttam Kumaran: they’re all, like, not insignificant. Like, some of the stuff is just docs or markdown files, but some of it is real features.
151 00:19:15.410 ⇒ 00:19:28.380 Uttam Kumaran: So then my question to you is, like, okay, how do you support that, like, as a platform? I don’t think that’s necessarily, like, gonna be a client thing, but as now. The second piece, which is actually gonna be client-facing, is,
152 00:19:29.310 ⇒ 00:19:33.789 Uttam Kumaran: like, Clarence is gonna start to work with us in a deeper way.
153 00:19:33.940 ⇒ 00:19:37.360 Samuel Roberts: Okay. Basically created a fork of open work.
154 00:19:37.360 ⇒ 00:19:42.230 Uttam Kumaran: That is something that we’re gonna actually try and, like, pitch to clients.
155 00:19:42.380 ⇒ 00:19:45.319 Samuel Roberts: Okay. And so a lot of this is in that open work.
156 00:19:45.510 ⇒ 00:19:47.260 Uttam Kumaran: integration plan.
157 00:19:47.840 ⇒ 00:19:55.330 Uttam Kumaran: And this is where, like, that’s probably where I immediately, in addition to just the, like, platform stability.
158 00:19:55.700 ⇒ 00:19:58.209 Uttam Kumaran: that’s where I think we’re gonna pro-
159 00:19:58.370 ⇒ 00:20:04.830 Uttam Kumaran: once we get this stood up, like, more than likely, some people may stop using cursor and transition
160 00:20:05.170 ⇒ 00:20:08.440 Uttam Kumaran: OpenWork. I don’t know if you’ve, like, seen OpenWork at all, or, like.
161 00:20:08.440 ⇒ 00:20:15.129 Samuel Roberts: I mean, I only saw some of this stuff today when I was getting the refund. It’s basically, like, Claude Co-Work, it’s like a clone of that. Yeah.
162 00:20:15.430 ⇒ 00:20:22.620 Uttam Kumaran: And… It’s a… it’s a… it’s a clone of it, but it’s like…
163 00:20:23.050 ⇒ 00:20:25.630 Uttam Kumaran: It’s something, like, composable, like, we can edit, and so…
164 00:20:25.630 ⇒ 00:20:26.409 Samuel Roberts: Sure, sure.
165 00:20:26.410 ⇒ 00:20:36.549 Uttam Kumaran: thinking about, basically, because one thing we’re finding is that, like, not a lot of our clients at the moment are going to do what I decided, which is, like, non-technical users are going to use cursor.
166 00:20:36.550 ⇒ 00:20:37.130 Samuel Roberts: Yeah, yeah, yeah.
167 00:20:37.130 ⇒ 00:20:42.929 Uttam Kumaran: Right? Like, that was a little bit of a stretch, so a lot… so we’re thinking about… we’re basically…
168 00:20:44.250 ⇒ 00:20:52.199 Uttam Kumaran: We’ve confirmed that the alpha in what we did internally was actually less of the interface, although previously
169 00:20:52.530 ⇒ 00:20:58.849 Uttam Kumaran: no interface, is actually all of the context availability and the SOP work.
170 00:20:59.550 ⇒ 00:21:02.390 Uttam Kumaran: That’s the alpha, and so really…
171 00:21:02.560 ⇒ 00:21:04.730 Uttam Kumaran: We want to go do that for folks.
172 00:21:05.070 ⇒ 00:21:05.810 Samuel Roberts: Totally.
173 00:21:05.810 ⇒ 00:21:12.650 Uttam Kumaran: Like, we want to go do kind of what we did for Andy, less of what we did for Lilo.
174 00:21:12.650 ⇒ 00:21:13.270 Samuel Roberts: Yeah.
175 00:21:13.270 ⇒ 00:21:14.940 Uttam Kumaran: Like, more on this, like.
176 00:21:15.050 ⇒ 00:21:22.900 Uttam Kumaran: kind of building the knowledge layer, you know, like, building the context layer. And then there’s some layers on top of that, like, we can think about, but…
177 00:21:23.190 ⇒ 00:21:30.630 Uttam Kumaran: this is the thing that I found was… is so incredible for our business, that, like, we’re… we’re basically gonna try and pitch open work
178 00:21:30.740 ⇒ 00:21:34.559 Uttam Kumaran: as, like, the UI on top of, like, a context layer.
179 00:21:34.680 ⇒ 00:21:39.230 Uttam Kumaran: Cursor is probably just… is just a bit more advanced, but similar deal, like…
180 00:21:39.230 ⇒ 00:21:39.880 Samuel Roberts: Yeah, yeah.
181 00:21:39.880 ⇒ 00:21:41.310 Uttam Kumaran: You know? So…
182 00:21:41.310 ⇒ 00:22:00.060 Samuel Roberts: I was just… I was just saying this, I think, I actually had a call with Brielle earlier, and I was telling, like, the different UIs, and, like, I’ve been using Codex a little bit, like, the app, just on a few things, instead of Cursor, and there’s definitely something about that, like, not everything needs an IDE, not everything needs all this other stuff that Cursor has, and so this… now I’m looking at this, I’m like, oh, yeah.
183 00:22:00.060 ⇒ 00:22:02.030 Samuel Roberts: CloudWare. Like, this is… that’s where it’s going, I feel.
184 00:22:02.030 ⇒ 00:22:07.660 Uttam Kumaran: Yes. And so Clarence’s point, and Clarence comes from more of the business world, is like, we’re never gonna be able to sell cursor.
185 00:22:07.840 ⇒ 00:22:08.600 Samuel Roberts: No, no, I agree.
186 00:22:08.600 ⇒ 00:22:10.189 Uttam Kumaran: with him, but he’s like…
187 00:22:10.450 ⇒ 00:22:24.969 Uttam Kumaran: this… he’s been doing a bunch of demos with OpenWork, and it’s been working, and then he put OpenWork on top of our context layer, and he’s like, holy shit, it’s, like, really, really, really good. And then, again, OpenWork is a UI on top of skills, on top of commands, so…
188 00:22:25.560 ⇒ 00:22:27.810 Uttam Kumaran: So eventually, like, that’s… I think…
189 00:22:27.940 ⇒ 00:22:32.569 Uttam Kumaran: I think what I’m kind of thinking about is, like, our platform is going to
190 00:22:32.700 ⇒ 00:22:38.299 Uttam Kumaran: be… in terms of Brainforge stuff, I think our platform will be… you’ll log into OpenWork.
191 00:22:38.390 ⇒ 00:22:42.419 Samuel Roberts: There will be tools and stuff available, like tools and other UIs.
192 00:22:42.600 ⇒ 00:22:45.470 Uttam Kumaran: But that will be the interface, actually, on the platform.
193 00:22:45.890 ⇒ 00:22:46.500 Samuel Roberts: Yeah.
194 00:22:46.650 ⇒ 00:22:53.650 Uttam Kumaran: In addition to, like, all the widgets and things that we’ve built, but, like, but also we are planning on using that for demos.
195 00:22:54.230 ⇒ 00:22:54.930 Uttam Kumaran: So that’s, like.
196 00:22:54.930 ⇒ 00:22:55.740 Samuel Roberts: Mmm.
197 00:22:55.740 ⇒ 00:22:58.609 Uttam Kumaran: That’s the one thing that I probably need help with as well, is, like.
198 00:22:58.870 ⇒ 00:23:08.169 Uttam Kumaran: Because he not only shipped it as, like, a web app, but it’s… he also is dipping it as, like, as a Apple application, like, he got it signed.
199 00:23:08.560 ⇒ 00:23:08.950 Samuel Roberts: Oh, nice.
200 00:23:08.950 ⇒ 00:23:13.360 Uttam Kumaran: And so there’s some things that once we get it shipped as a web app, we can talk about that. Yeah.
201 00:23:15.030 ⇒ 00:23:19.109 Uttam Kumaran: But… so that’s the biggest kind of thing there, and then…
202 00:23:19.890 ⇒ 00:23:24.809 Uttam Kumaran: So the open work is something that I think I want to kind of get you, Ms. Safa.
203 00:23:25.580 ⇒ 00:23:28.549 Uttam Kumaran: Aware of, and, like, poking at that plan.
204 00:23:28.660 ⇒ 00:23:31.560 Uttam Kumaran: And maybe even, like, taking pieces of it, like.
205 00:23:31.560 ⇒ 00:23:32.270 Samuel Roberts: Yeah.
206 00:23:32.270 ⇒ 00:23:38.259 Uttam Kumaran: And basically finding out, like, how we can get that stood up. And then, it’s sort of just, again, just helping me, like.
207 00:23:38.260 ⇒ 00:23:40.539 Samuel Roberts: ship faster. Like, I think that’s where…
208 00:23:40.560 ⇒ 00:23:44.120 Uttam Kumaran: even when you were working the HubSpot, I think I’m like, just give… I’ll take it.
209 00:23:44.240 ⇒ 00:23:52.000 Samuel Roberts: Instead, it’s like, I need your… you and your brain to focus on… it’s like, sort of like what we talked about 3 months ago, or whatever we talked, right, or 4 months ago. Yeah.
210 00:23:52.000 ⇒ 00:23:58.860 Uttam Kumaran: Which was, like, the limitation… Is now gonna be more on, like… I’m the first one…
211 00:23:59.270 ⇒ 00:24:03.130 Uttam Kumaran: In the business to get the bug for, like, shipping all this stuff.
212 00:24:03.430 ⇒ 00:24:07.559 Uttam Kumaran: Bryle is now doing it, like, it will start to grow.
213 00:24:07.680 ⇒ 00:24:18.930 Uttam Kumaran: And so the platform, the level of our testing and CI needs to, like, kind of, like, speak for itself, so that… because otherwise, we’re going to become the bottleneck on reviews.
214 00:24:19.520 ⇒ 00:24:20.010 Samuel Roberts: Yes.
215 00:24:20.010 ⇒ 00:24:27.490 Uttam Kumaran: So that’s really, like, where I think using your brain on that, like, how do we create different environments for people to be able to ship more things?
216 00:24:28.000 ⇒ 00:24:33.809 Uttam Kumaran: Because it’s just gonna increase, like, I have a feeling. Even just the last two weeks, so…
217 00:24:36.080 ⇒ 00:24:48.210 Samuel Roberts: Definitely. Yeah, okay. I mean, I’ve been sort of mulling that over in terms of, like, testing environments and stuff, trying to think through that, like, because I think that is exactly, you know, you said, like, you’re just messing with environment variables, trying to get things right, and it’s like…
218 00:24:48.400 ⇒ 00:24:53.540 Samuel Roberts: There’s not the… you can’t just feed that to the AI, necessarily, yet, you know?
219 00:24:53.650 ⇒ 00:24:58.579 Samuel Roberts: But if we can streamline that, get that going, people can definitely contribute a lot more.
220 00:24:59.530 ⇒ 00:25:01.450 Uttam Kumaran: Yeah, I think people are sort of…
221 00:25:02.290 ⇒ 00:25:04.880 Uttam Kumaran: Like, talking about it, kind of like…
222 00:25:08.520 ⇒ 00:25:11.739 Uttam Kumaran: But kind of talking about a, like, harness engineering.
223 00:25:11.740 ⇒ 00:25:13.730 Samuel Roberts: Yes, yes, I’ve seen that recently, yeah.
224 00:25:13.730 ⇒ 00:25:18.140 Uttam Kumaran: So I’ll send you some… I’ll send you some things.
225 00:25:19.270 ⇒ 00:25:21.100 Uttam Kumaran: that I’ve been, like, reading on that, like.
226 00:25:21.100 ⇒ 00:25:21.909 Samuel Roberts: Yeah, right? Yeah.
227 00:25:22.430 ⇒ 00:25:25.190 Uttam Kumaran: this, and then there’s also cursor…
228 00:25:26.470 ⇒ 00:25:31.489 Samuel Roberts: I think I saw this in an OpenAI email. Yeah, okay, I know, I… yeah, so the first one.
229 00:25:31.660 ⇒ 00:25:35.749 Samuel Roberts: I saw… I didn’t… It was over the weekend, so I didn’t get to dig into it, but I was like, yes.
230 00:25:36.580 ⇒ 00:25:38.359 Samuel Roberts: Driving codebases, yeah.
231 00:25:41.030 ⇒ 00:25:41.840 Uttam Kumaran: So…
232 00:25:41.840 ⇒ 00:25:43.700 Samuel Roberts: Like, this is where…
233 00:25:43.700 ⇒ 00:25:47.350 Uttam Kumaran: Like, we’re… we’re gonna be here very soon, and so…
234 00:25:47.350 ⇒ 00:25:47.950 Samuel Roberts: Yeah.
235 00:25:49.320 ⇒ 00:25:55.870 Uttam Kumaran: biasing towards a world where people are kicking off multiple VMs, working on different parts of tasks.
236 00:25:56.280 ⇒ 00:26:02.480 Uttam Kumaran: And just believing that that’s gonna be the future is sort of, like, where I kind of would love to have your brain on, like.
237 00:26:02.670 ⇒ 00:26:06.250 Uttam Kumaran: Like, and I don’t think it’s, like, reinventing, it’s actually just, like.
238 00:26:06.380 ⇒ 00:26:10.660 Uttam Kumaran: Probably have to go find, like, what the best practices are from, like, bigger companies on, like.
239 00:26:10.660 ⇒ 00:26:11.380 Samuel Roberts: Yeah.
240 00:26:11.380 ⇒ 00:26:17.530 Uttam Kumaran: environment and CI, And then, you know, like, coming up with a framework for that, but that…
241 00:26:17.680 ⇒ 00:26:24.299 Uttam Kumaran: That’s where I think our… like, I can now… I feel comfortable now shipping features.
242 00:26:24.510 ⇒ 00:26:25.020 Samuel Roberts: Right.
243 00:26:25.020 ⇒ 00:26:29.569 Uttam Kumaran: And so I can take that, If you can… take, like…
244 00:26:30.380 ⇒ 00:26:32.610 Samuel Roberts: How do we make sure that they…
245 00:26:32.980 ⇒ 00:26:37.960 Uttam Kumaran: are able to be run, whether through cursor, through agents, and, like, the CI works.
246 00:26:38.580 ⇒ 00:26:44.490 Uttam Kumaran: That’s… and then… where I’m gonna even challenge this further is I’m… I’m,
247 00:26:44.850 ⇒ 00:26:51.899 Uttam Kumaran: I’m thinking about working on an even larger project, basically around doing the same thing, except for data.
248 00:26:52.670 ⇒ 00:26:53.380 Samuel Roberts: Okay.
249 00:26:53.540 ⇒ 00:26:56.110 Uttam Kumaran: So one of the things that’s difficult right now is that
250 00:26:56.320 ⇒ 00:27:04.200 Uttam Kumaran: only the AI full… only full-stack engineers are getting, like, all the love in terms of, like…
251 00:27:04.520 ⇒ 00:27:06.760 Uttam Kumaran: Using AI to do their work.
252 00:27:07.020 ⇒ 00:27:09.940 Samuel Roberts: Data people are not getting that much love.
253 00:27:10.140 ⇒ 00:27:15.169 Uttam Kumaran: And so, basically, I’m like, what did we learn from me trying to figure out the AI side?
254 00:27:15.440 ⇒ 00:27:20.709 Uttam Kumaran: that we can bring. And so, really, it’s like, I wrote down some docs, which is, like.
255 00:27:21.530 ⇒ 00:27:23.930 Uttam Kumaran: Basically, thinking about all the tasks
256 00:27:24.250 ⇒ 00:27:29.009 Uttam Kumaran: Around AI… around data engineering, analytics engineering, and analysis.
257 00:27:29.500 ⇒ 00:27:37.310 Uttam Kumaran: And then trying to basically think about, like, what are the workflows that we need to enable
258 00:27:37.510 ⇒ 00:27:41.449 Uttam Kumaran: And, like, what is the CI step needed for every type of workflow, right?
259 00:27:41.450 ⇒ 00:27:42.570 Samuel Roberts: Okay, yeah.
260 00:27:42.570 ⇒ 00:27:57.540 Uttam Kumaran: let’s say you were able to… you’re building a dbt model, okay, like, well, there needs to be, like, a bunch of CI that runs on checking, like, hey, this column got added, it impacted rows this way. Oh, okay, yeah. What is it on… what is… but what is it for analysts, too? Like, you know, like…
261 00:27:57.860 ⇒ 00:28:04.240 Uttam Kumaran: Let’s say a client asks a question, do the analysts literally just put the question into chat, and then it, like, works on a plan for them?
262 00:28:04.390 ⇒ 00:28:05.140 Samuel Roberts: Right.
263 00:28:05.140 ⇒ 00:28:12.069 Uttam Kumaran: So I sort of wrote this doc as, like, basically, like, yeah, what is the hardness for, like, an analyst, right? Like, there’s an intake question.
264 00:28:12.380 ⇒ 00:28:19.049 Uttam Kumaran: then there’s all these different things, and then how do they build a deck? So it’s all kind of coming together to sort of align with, like.
265 00:28:20.170 ⇒ 00:28:23.740 Uttam Kumaran: how do I make sure that the…
266 00:28:23.930 ⇒ 00:28:31.580 Uttam Kumaran: platform and cursor and agents support any type of work that happened at Brainforge, starting with delivery.
267 00:28:31.830 ⇒ 00:28:40.210 Uttam Kumaran: And delivery, of course, includes everything on our AI team, and from our data team, and from our strategy team.
268 00:28:40.470 ⇒ 00:28:43.519 Uttam Kumaran: I’m less worried about sales.
269 00:28:44.110 ⇒ 00:28:56.720 Uttam Kumaran: marketing, these guys, because they’ll start to pick it up for themselves, but delivery is our largest cost center, and so I’m like, okay, so… and also the data team does not currently use a lot of cursor, mainly because
270 00:28:57.130 ⇒ 00:28:59.590 Uttam Kumaran: like, it’s a little bit difficult, right?
271 00:28:59.590 ⇒ 00:29:00.020 Samuel Roberts: That’s sort of like…
272 00:29:00.020 ⇒ 00:29:02.350 Uttam Kumaran: like, what I’m thinking about is, like, how do we build…
273 00:29:02.770 ⇒ 00:29:10.550 Uttam Kumaran: like, how do we just… what is, like, a rollout plan of a bunch of, like, AI harnessing for a variety of data agents?
274 00:29:10.750 ⇒ 00:29:12.790 Uttam Kumaran: You know… And then…
275 00:29:12.790 ⇒ 00:29:13.450 Samuel Roberts: Okay.
276 00:29:13.450 ⇒ 00:29:19.249 Uttam Kumaran: I read a lot about, like, these, like, self-learning data agents and things like that, but really, I’m, like, trying to think about
277 00:29:19.690 ⇒ 00:29:23.729 Uttam Kumaran: Internal teams can run a question-to-evidence Insight to deck.
278 00:29:24.450 ⇒ 00:29:28.860 Uttam Kumaran: Like, with outputs, right? And, like, how does this happen at Brainforge?
279 00:29:29.000 ⇒ 00:29:32.620 Uttam Kumaran: And how does this happen? Like, how does Amber walk in and say, I have this question.
280 00:29:32.810 ⇒ 00:29:35.680 Uttam Kumaran: And she’s able to get a deck, fully baked deck.
281 00:29:36.380 ⇒ 00:29:43.429 Uttam Kumaran: like, that’s high quality within an hour, versus it taking, like, 2 days, a few hours, right? So that’s the things I’m working on, so…
282 00:29:43.930 ⇒ 00:29:48.770 Samuel Roberts: all of what I’m gonna say is, like, CI is gonna be even a bigger part of this, and it’s gonna be…
283 00:29:48.770 ⇒ 00:29:52.510 Uttam Kumaran: It’s gonna look a lot different than what’s traditional with
284 00:29:52.780 ⇒ 00:29:56.030 Uttam Kumaran: software engineering, like, it’s not exactly, like, linting and, like.
285 00:29:56.190 ⇒ 00:29:58.329 Samuel Roberts: running NPM, it’s like…
286 00:29:58.480 ⇒ 00:30:03.040 Uttam Kumaran: Checking columns, and like… Running, like, an agent that
287 00:30:03.270 ⇒ 00:30:05.979 Uttam Kumaran: Looks at our, like, analysis standards, so, like.
288 00:30:06.090 ⇒ 00:30:08.999 Uttam Kumaran: CI may look a little bit different for this kind of data work.
289 00:30:10.200 ⇒ 00:30:10.840 Samuel Roberts: Okay.
290 00:30:10.940 ⇒ 00:30:11.969 Samuel Roberts: Yeah, that’s…
291 00:30:12.080 ⇒ 00:30:17.199 Samuel Roberts: I can see how it’s different, or I can see that it’s different, I don’t necessarily know the ins and outs, so this will be interesting to look at, yeah.
292 00:30:17.200 ⇒ 00:30:17.890 Uttam Kumaran: Yeah.
293 00:30:18.430 ⇒ 00:30:24.619 Uttam Kumaran: So that’s where, like, I’m just gonna follow… I’m gonna try to follow, like, whatever best practices you put in place.
294 00:30:25.180 ⇒ 00:30:32.619 Uttam Kumaran: because again, we have GitHub actions, we have our traditional deployment-related CI steps, we have pre-commits.
295 00:30:33.840 ⇒ 00:30:38.110 Uttam Kumaran: So, what is the… what is the array of options that we’re doing, you know?
296 00:30:38.370 ⇒ 00:30:39.060 Samuel Roberts: Yeah.
297 00:30:41.710 ⇒ 00:30:42.570 Samuel Roberts: Okay.
298 00:30:44.710 ⇒ 00:30:45.420 Samuel Roberts: Okay.
299 00:30:49.420 ⇒ 00:30:56.530 Samuel Roberts: Okay, yeah, I think, yeah, I mean, I was actually… I was digging into some of the stuff, because things weren’t running right on GitHub, and I’m wondering, like, is… yeah, well, we’ll dig into that, but GitHub.
300 00:30:56.530 ⇒ 00:31:02.570 Uttam Kumaran: I mean, part of it is, like, every time we said, like, yeah, but how can we also use AI to, like, build that test suite up and things like that?
301 00:31:02.570 ⇒ 00:31:09.939 Samuel Roberts: Yeah, oh, definitely, definitely. I have, yeah, no intention of writing that, believe me. It’s more just, like, thinking through what… where to point it, and what to…
302 00:31:11.820 ⇒ 00:31:14.469 Samuel Roberts: What to specify, yeah. Okay, cool.
303 00:31:16.320 ⇒ 00:31:18.160 Samuel Roberts: It’s good.
304 00:31:20.540 ⇒ 00:31:25.119 Samuel Roberts: Where was the… the open work fork? I know I saw something earlier, but now I’m like…
305 00:31:25.120 ⇒ 00:31:30.410 Uttam Kumaran: Oh, it’s in the, it’s in the, well, no, it’s under app.
306 00:31:30.410 ⇒ 00:31:31.110 Samuel Roberts: I hope now.
307 00:31:31.310 ⇒ 00:31:37.150 Samuel Roberts: It’s under apps. Okay, that’s why I was just looking to see where the… okay, cool, cool, I will… I’ll dig into that. Cool, that’s sweet, I want to play with that.
308 00:31:37.520 ⇒ 00:31:39.510 Samuel Roberts: And then there was another thought there…
309 00:31:41.410 ⇒ 00:31:47.029 Samuel Roberts: Yeah, okay, I think… did Pranav grab time tomorrow? I don’t think so, but I’ll make sure to get some.
310 00:31:47.160 ⇒ 00:31:53.059 Samuel Roberts: In terms of the testing for the platform to get that in there. And then, yeah, just thinking about the testing harness in general, and…
311 00:31:53.300 ⇒ 00:31:58.029 Samuel Roberts: Okay. Is that… what you were just showing me, was that a PR open, or a draft PR or something?
312 00:31:58.320 ⇒ 00:32:01.450 Uttam Kumaran: Yes, that is also just draft PR sitting there.
313 00:32:01.630 ⇒ 00:32:04.319 Samuel Roberts: Okay, cool. Try to take a look at that, because that would be…
314 00:32:04.670 ⇒ 00:32:07.809 Samuel Roberts: good to get a full context of where it’s going, because I don’t… like I said, I don’t…
315 00:32:08.020 ⇒ 00:32:11.520 Samuel Roberts: know the data side as much, and what would CI look like for that?
316 00:32:12.740 ⇒ 00:32:13.530 Samuel Roberts: Okay.
317 00:32:15.300 ⇒ 00:32:16.990 Samuel Roberts: Cool. Alright.
318 00:32:17.350 ⇒ 00:32:20.669 Samuel Roberts: There’s another thought, I forgot what it was, but…
319 00:32:21.570 ⇒ 00:32:23.540 Samuel Roberts: It was much earlier in the day, so it’s fine.
320 00:32:25.490 ⇒ 00:32:29.800 Samuel Roberts: Alright, oh, crap, it’s 5. Okay, I may have to run.
321 00:32:29.800 ⇒ 00:32:30.430 Uttam Kumaran: Okay.
322 00:32:30.790 ⇒ 00:32:32.250 Samuel Roberts: To relieve the babysitter.
323 00:32:33.080 ⇒ 00:32:34.659 Samuel Roberts: Anything else, I guess?
324 00:32:35.260 ⇒ 00:32:36.850 Uttam Kumaran: I think that’s it.
325 00:32:37.070 ⇒ 00:32:40.970 Samuel Roberts: Okay, cool. I’ll digest this a little bit,
326 00:32:41.310 ⇒ 00:32:43.069 Samuel Roberts: Maybe you can catch up later this week, too.
327 00:32:43.270 ⇒ 00:32:44.520 Uttam Kumaran: Okay, okay, perfect.
328 00:32:44.830 ⇒ 00:32:46.080 Samuel Roberts: Cool. Thank you, dude.
329 00:32:46.260 ⇒ 00:32:47.010 Samuel Roberts: Alrighty.
330 00:32:47.140 ⇒ 00:32:47.720 Samuel Roberts: Have a good one.
331 00:32:47.950 ⇒ 00:32:49.039 Uttam Kumaran: That isn’t buy.
332 00:32:49.040 ⇒ 00:32:50.080 Samuel Roberts: Alrighty, bye.