Meeting Title: AI Team Retro Date: 2025-08-08 Meeting participants: Mustafa Raja, Sam Roberts, Casie Aviles, Uttam Kumaran
WEBVTT
1 00:01:39.510 ⇒ 00:01:40.590 Sam Roberts: Hello!
2 00:01:42.840 ⇒ 00:01:44.039 Mustafa Raja: Hey! How are you?
3 00:01:44.970 ⇒ 00:01:46.889 Sam Roberts: Doing all right. How about yourself?
4 00:01:47.530 ⇒ 00:01:48.900 Mustafa Raja: Yeah, I’m doing good.
5 00:01:50.380 ⇒ 00:01:52.460 Sam Roberts: That’s too bad. About the timeout stuff.
6 00:01:52.460 ⇒ 00:02:00.560 Mustafa Raja: Yeah, I I implemented the solution that we discussed. And I was like, it’s still timing out, why is that.
7 00:02:01.037 ⇒ 00:02:02.469 Sam Roberts: Oh! And then.
8 00:02:02.470 ⇒ 00:02:05.210 Mustafa Raja: I checked the logs
9 00:02:05.790 ⇒ 00:02:12.739 Mustafa Raja: and it said, It’s from Cloudflare. And I said, Oh, that’s weird! Why is it from Cloudflare?
10 00:02:12.980 ⇒ 00:02:13.790 Mustafa Raja: And then.
11 00:02:14.690 ⇒ 00:02:15.590 Sam Roberts: Yeah.
12 00:02:15.590 ⇒ 00:02:19.559 Mustafa Raja: Yeah, it should. It should have been in the docs, though in an.
13 00:02:19.560 ⇒ 00:02:21.599 Sam Roberts: You would think so. Yeah. But you know.
14 00:02:21.600 ⇒ 00:02:26.760 Mustafa Raja: It’s never! It’s it’s it isn’t even mentioned there it should be mentioned. There.
15 00:02:27.480 ⇒ 00:02:29.219 Sam Roberts: Oh, yeah, I really should be.
16 00:02:29.220 ⇒ 00:02:35.590 Mustafa Raja: Yeah, that sucks. But it’s okay. At least we were able to pinpoint the problem right?
17 00:02:36.910 ⇒ 00:02:43.120 Sam Roberts: Yeah, yeah, I mean, it’s definitely a a path forward at least.
18 00:02:43.120 ⇒ 00:02:43.670 Mustafa Raja: Yeah.
19 00:02:43.670 ⇒ 00:02:46.410 Sam Roberts: It’s definitely and definitely good to know now. Certainly, too.
20 00:02:46.410 ⇒ 00:02:51.080 Mustafa Raja: Yeah, yeah, yeah, we know, we know for the future. If if it comes again.
21 00:02:51.580 ⇒ 00:02:52.880 Sam Roberts: Yeah, exactly.
22 00:02:53.300 ⇒ 00:02:54.520 Mustafa Raja: Hey, Jesse, how are you?
23 00:02:57.010 ⇒ 00:02:59.769 Casie Aviles: A, yeah, I’m fine.
24 00:03:03.450 ⇒ 00:03:04.890 Mustafa Raja: Any plans for the weekend.
25 00:03:06.400 ⇒ 00:03:10.040 Casie Aviles: I’m just gonna try to rest as much as I can.
26 00:03:10.040 ⇒ 00:03:10.600 Mustafa Raja: It’s the best.
27 00:03:10.600 ⇒ 00:03:12.090 Casie Aviles: This week has been crazy.
28 00:03:13.130 ⇒ 00:03:14.220 Mustafa Raja: Steve.
29 00:03:18.680 ⇒ 00:03:20.821 Sam Roberts: Yeah, it was winter. Nice.
30 00:03:21.790 ⇒ 00:03:22.380 Mustafa Raja: So how’s.
31 00:03:22.380 ⇒ 00:03:23.029 Sam Roberts: Especially after.
32 00:03:23.030 ⇒ 00:03:25.440 Mustafa Raja: How’s your week been, Sam?
33 00:03:26.339 ⇒ 00:03:39.760 Sam Roberts: I was been pretty good, you know. I’m I’m feeling more comfortable with everything my head around stuff and all the processes, and set it on some like client stuff. So I’m starting to get my bearings a little bit better.
34 00:03:41.510 ⇒ 00:03:45.150 Sam Roberts: yeah, it’s been. It’s been good, you know. I’m enjoying it.
35 00:03:46.460 ⇒ 00:03:47.550 Sam Roberts: It’s cool stuff.
36 00:03:52.880 ⇒ 00:03:57.920 Sam Roberts: Yeah, I’m actually playing around. So yesterday they released Gpt 5.
37 00:03:57.920 ⇒ 00:03:58.770 Sam Roberts: Oh, yeah.
38 00:03:58.770 ⇒ 00:03:59.160 Sam Roberts: So.
39 00:03:59.160 ⇒ 00:03:59.980 Mustafa Raja: Good, actually.
40 00:04:02.140 ⇒ 00:04:03.760 Mustafa Raja: It’s super good. I felt.
41 00:04:03.760 ⇒ 00:04:12.790 Sam Roberts: Oh, yeah, yeah, I I yeah. So I I saw the demo and was like, well, I had talked. I mentioned that I wanted to try to put a plan together to migrate
42 00:04:12.930 ⇒ 00:04:15.569 Sam Roberts: the front end and back end of the platform.
43 00:04:16.130 ⇒ 00:04:20.360 Sam Roberts: And I was like, let’s just see if Gpt 5 can take a swing at it. And so
44 00:04:20.490 ⇒ 00:04:24.810 Sam Roberts: it’s kind of just running now, and it’s it’s migrating things pretty well, hey?
45 00:04:24.810 ⇒ 00:04:25.540 Mustafa Raja: Your time.
46 00:04:25.650 ⇒ 00:04:26.160 Mustafa Raja: Nice.
47 00:04:26.160 ⇒ 00:04:27.540 Uttam Kumaran: Hey! How are you?
48 00:04:28.200 ⇒ 00:04:30.060 Sam Roberts: Good! Good! How are you doing.
49 00:04:30.940 ⇒ 00:04:31.940 Uttam Kumaran: Good.
50 00:04:35.330 ⇒ 00:04:43.289 Uttam Kumaran: Cool. Yeah. What do you guys think you guys watch the the live stream? Yesterday.
51 00:04:44.170 ⇒ 00:04:48.640 Mustafa Raja: I watched it for like 20 min I did, and did a little sneak peek.
52 00:04:48.640 ⇒ 00:04:52.000 Uttam Kumaran: Kind of boring. I fast forward through a lot of it.
53 00:04:52.720 ⇒ 00:04:59.100 Sam Roberts: Yeah, I I forget when I jumped in, but I I went back a little bit, and then 2 x through the kind of demo of getting cursor.
54 00:05:00.618 ⇒ 00:05:08.560 Sam Roberts: But that’s where I’ll be touching it, most probably, I imagine. And so that’s what I was just saying. It’s it’s been pretty good today and a little bit last night just
55 00:05:09.694 ⇒ 00:05:11.590 Sam Roberts: just chugging through stuff.
56 00:05:11.800 ⇒ 00:05:14.600 Sam Roberts: It’s not the fastest, but it’s pretty good.
57 00:05:14.810 ⇒ 00:05:15.943 Uttam Kumaran: Okay. Cool.
58 00:05:17.780 ⇒ 00:05:19.439 Mustafa Raja: I wonder if we can.
59 00:05:19.831 ⇒ 00:05:22.570 Uttam Kumaran: I didn’t get the update yet. So.
60 00:05:23.400 ⇒ 00:05:24.380 Sam Roberts: Oh, yeah.
61 00:05:24.870 ⇒ 00:05:29.789 Mustafa Raja: I wonder if we can deploy it to our azure endpoint.
62 00:05:30.950 ⇒ 00:05:32.880 Uttam Kumaran: Yes, if you can.
63 00:05:33.920 ⇒ 00:05:34.470 Mustafa Raja: Okay.
64 00:05:35.890 ⇒ 00:05:36.800 Uttam Kumaran: I didn’t.
65 00:05:38.300 ⇒ 00:05:41.580 Uttam Kumaran: Yeah, I still don’t see it in my like chatgpt desktop.
66 00:05:41.580 ⇒ 00:05:45.729 Sam Roberts: Yeah, I don’t see it there, either, but in cursor. When I I reboot, I read, you know.
67 00:05:45.890 ⇒ 00:05:50.470 Sam Roberts: every once in a while I realize curse has been open for days and has an update. So I finally ran it, and
68 00:05:50.760 ⇒ 00:05:51.520 Uttam Kumaran: Alright!
69 00:05:52.090 ⇒ 00:05:54.679 Mustafa Raja: It’s in the. It’s updated in the web, though.
70 00:05:55.370 ⇒ 00:05:55.970 Sam Roberts: Cool.
71 00:06:00.140 ⇒ 00:06:01.420 Uttam Kumaran: Oh, it is really, okay.
72 00:06:01.420 ⇒ 00:06:02.010 Mustafa Raja: Yeah.
73 00:06:06.558 ⇒ 00:06:12.139 Mustafa Raja: I couldn’t find older models actually on the web. It’s only 5 5 thinking.
74 00:06:12.140 ⇒ 00:06:13.779 Uttam Kumaran: On the web. Wait! Hold on!
75 00:06:13.960 ⇒ 00:06:16.690 Uttam Kumaran: Oh, I see it’s on my team account.
76 00:06:17.710 ⇒ 00:06:18.600 Uttam Kumaran: It’s on my personal.
77 00:06:18.600 ⇒ 00:06:19.870 Sam Roberts: Oh, interesting!
78 00:06:23.230 ⇒ 00:06:23.980 Sam Roberts: Huh!
79 00:06:26.990 ⇒ 00:06:31.350 Uttam Kumaran: So I guess it’ll it’ll get there soon, probably next week.
80 00:06:33.090 ⇒ 00:06:36.300 Uttam Kumaran: Okay, I’ll I’ll I’ll try it today for some stuff.
81 00:06:40.110 ⇒ 00:06:51.510 Sam Roberts: Yeah, I was just mentioned as you were hopping on, that I was gonna start putting like a plan together to migrate the platform to next and have it. And I was like, let’s just see if Gpt 5 can do it, and it’s because I I ran a little.
82 00:06:51.510 ⇒ 00:06:52.520 Sam Roberts: Oh.
83 00:06:53.350 ⇒ 00:07:05.400 Sam Roberts: and I was like, let me let me back up. So I was like, instead of just giving it each repo I was like, here are the 2 repos. I want to combine it and I was doing this like just to see if cursor could do it.
84 00:07:05.590 ⇒ 00:07:10.619 Sam Roberts: And it did all right before I ran it with Gpt. 5, but the styling was all whack.
85 00:07:12.330 ⇒ 00:07:15.910 Sam Roberts: and so they said like, Oh, Gpt is like 5 is so good with the
86 00:07:16.070 ⇒ 00:07:24.510 Sam Roberts: styling, and it’s a great designer and stuff, and I was like, well, maybe it’ll be better at hopping all that stuff over, and so far it’s looking pretty good still chugging away on some stuff. But
87 00:07:27.290 ⇒ 00:07:29.669 Sam Roberts: it’s yeah. It’s kind of impressive.
88 00:07:30.980 ⇒ 00:07:31.930 Uttam Kumaran: Nice.
89 00:07:34.760 ⇒ 00:07:45.380 Uttam Kumaran: Yeah, I was just thinking, even this morning, like, on our conversation about copilot kit, like a lot of our agents in the platform are actually just gonna be simple.
90 00:07:45.610 ⇒ 00:07:48.499 Uttam Kumaran: like, basic, just like prompt.
91 00:07:49.480 ⇒ 00:07:50.180 Sam Roberts: Right.
92 00:07:50.710 ⇒ 00:07:56.770 Uttam Kumaran: So like, for example, 1 1 of the things is like on the left side of the platform. We have all those like
93 00:07:57.350 ⇒ 00:08:00.400 Uttam Kumaran: basically like single use agents.
94 00:08:01.020 ⇒ 00:08:01.650 Sam Roberts: Yeah.
95 00:08:02.480 ⇒ 00:08:10.519 Uttam Kumaran: Those are all. I think we should just do actually like the Ui right now, sucks like, they’re kind of unusable. It’s like.
96 00:08:10.520 ⇒ 00:08:10.990 Uttam Kumaran: yeah, that’s it.
97 00:08:13.245 ⇒ 00:08:15.500 Uttam Kumaran: Artifacts
98 00:08:15.920 ⇒ 00:08:25.939 Uttam Kumaran: like. And a lot of people aren’t gonna even know like how to use it. So you’ll need like some tutorials or something like we should totally move those over.
99 00:08:26.560 ⇒ 00:08:43.810 Sam Roberts: Yeah, yeah, that’s that’s I mean. So I I realized I had the co-pilot kit stuff. And I I kind of commented it out so that it wouldn’t break, as I was migrating things. But that should be the easiest place to drop it in, because if it is just pulling prompts from super base. It’ll be really easy to just drop those into copilot kit.
100 00:08:44.120 ⇒ 00:08:50.410 Sam Roberts: And at least, you know, just a simple. Are these all chats, or these have other things. These are all chats in the AI agents.
101 00:08:50.710 ⇒ 00:08:56.619 Sam Roberts: Yeah, they’re not like and doing anything else right now. But with copilot we could probably more easily add, like
102 00:08:57.460 ⇒ 00:09:01.439 Sam Roberts: something else to that, because it had a few other. It had, like a text area that had, like
103 00:09:01.600 ⇒ 00:09:03.480 Sam Roberts: smart autocomplete and stuff.
104 00:09:04.250 ⇒ 00:09:11.820 Uttam Kumaran: Well, yeah, basically, I want, I use so many different projects in my chat. Gbt, that I want to start using in the open.
105 00:09:12.040 ⇒ 00:09:13.180 Uttam Kumaran: Yeah.
106 00:09:13.430 ⇒ 00:09:18.980 Uttam Kumaran: And I think another thing that I wanna one thing that we talked about is like, how can we show
107 00:09:19.170 ⇒ 00:09:24.280 Uttam Kumaran: that the chat history from like anyone who’s been using.
108 00:09:25.150 ⇒ 00:09:25.930 Sam Roberts: Hmm.
109 00:09:26.240 ⇒ 00:09:27.089 Uttam Kumaran: You know.
110 00:09:28.120 ⇒ 00:09:28.910 Sam Roberts: Yes.
111 00:09:29.630 ⇒ 00:09:30.330 Sam Roberts: Yeah.
112 00:09:30.330 ⇒ 00:09:37.180 Uttam Kumaran: Because then that way, everybody in the company like, if you start, you can just go look at like what’s an example of what was some. What was someone doing before
113 00:09:37.550 ⇒ 00:09:38.250 Uttam Kumaran: interesting.
114 00:09:38.250 ⇒ 00:09:39.580 Sam Roberts: Okay, yeah, yeah.
115 00:09:39.580 ⇒ 00:09:42.580 Uttam Kumaran: It’s like a way to hack like sort of the
116 00:09:43.296 ⇒ 00:09:47.989 Uttam Kumaran: main main thing I’m seeing with people is they just have the kind of this cold start problem. So
117 00:09:48.160 ⇒ 00:09:53.670 Uttam Kumaran: examples and templates are a great way. And then the suggested replies are a great way,
118 00:09:59.100 ⇒ 00:10:10.119 Sam Roberts: Yeah, I think. I think that’s a good. That’s a good point. The culture there’s definitely has that like that blank page like, what do I do now? Like? I I have. I know I need something. I I but I don’t know the best way to like
119 00:10:10.500 ⇒ 00:10:12.420 Sam Roberts: yes, into it, with the, with the.
120 00:10:12.420 ⇒ 00:10:25.530 Uttam Kumaran: Or like. You’ve heard that it’s if you’ve heard that there’s something exists here, and you’re like, how do I even consider like, what could I use it for? Right? Because, like my for me, I’m like, Oh, you need to write a project then, like Use AI, and people probably are like
121 00:10:26.690 ⇒ 00:10:34.619 Uttam Kumaran: the gap between. Those things are probably pretty high for people like for us. We’d be like perfect. But for folks I don’t think they
122 00:10:35.320 ⇒ 00:10:41.969 Uttam Kumaran: they don’t. They don’t see like. Oh, I have to. I can use a prompt to then write another, prompt to then get that thing out so like.
123 00:10:42.230 ⇒ 00:10:42.590 Sam Roberts: Right.
124 00:10:42.590 ⇒ 00:10:44.770 Uttam Kumaran: Has to be some more training, for sure.
125 00:10:47.670 ⇒ 00:10:51.139 Sam Roberts: Yeah, yeah, I think those agents are a good.
126 00:10:51.650 ⇒ 00:10:56.466 Sam Roberts: a good place to start, because then we can. Yeah, I gotta figure out how copilot kit can handle
127 00:10:57.500 ⇒ 00:11:09.040 Sam Roberts: It’s like right now it it the I mean with just the demo I had the the chats were like carrying through when I clicked around, but then getting lost when I refresh. So I gotta figure out, if I can.
128 00:11:09.190 ⇒ 00:11:15.890 Sam Roberts: how easily I can plug something in because it’s built to do stuff with like lang graph, and a few other agents that have persistence kind of
129 00:11:16.610 ⇒ 00:11:17.750 Sam Roberts: right there.
130 00:11:19.480 ⇒ 00:11:25.699 Sam Roberts: So either getting something like that running on the back end that can just like persist the chat
131 00:11:25.990 ⇒ 00:11:30.210 Sam Roberts: pretty simply, or just, you know, saving things to super base and
132 00:11:30.400 ⇒ 00:11:32.760 Sam Roberts: pulling them up might be the easiest way to have it.
133 00:11:33.480 ⇒ 00:11:37.440 Uttam Kumaran: I gotta look into that a little bit. How that’ll interact with the n 8 n, stuff, too.
134 00:11:39.270 ⇒ 00:11:39.930 Uttam Kumaran: Yeah.
135 00:11:40.900 ⇒ 00:11:43.188 Uttam Kumaran: Cool. Okay, cool. Well, maybe we can just
136 00:11:44.380 ⇒ 00:11:46.060 Uttam Kumaran: like I I mean, we are. Just
137 00:11:46.590 ⇒ 00:11:48.860 Uttam Kumaran: take a quick look at the
138 00:11:49.350 ⇒ 00:11:55.960 Uttam Kumaran: board, and then if there’s anything we can discuss, but otherwise we can just talk about, you know, feedback from this week.
139 00:12:17.020 ⇒ 00:12:22.510 Uttam Kumaran: Okay, yeah. I mean, I think, probably that if I just to look at the AI team board, probably the biggest
140 00:12:23.775 ⇒ 00:12:27.244 Uttam Kumaran: we got a bunch of stuff done, I think.
141 00:12:28.300 ⇒ 00:12:32.080 Uttam Kumaran: still, that’s not been done is like the core client hubs.
142 00:12:33.600 ⇒ 00:12:35.110 Uttam Kumaran: You know, in a couple of
143 00:12:35.360 ⇒ 00:12:40.240 Uttam Kumaran: making sure that those are are ready to go. So I know, Casey, we’re doing like some discovery
144 00:12:42.140 ⇒ 00:12:45.790 Uttam Kumaran: and fit and kind of like fixing that, but otherwise I feel
145 00:12:46.720 ⇒ 00:12:51.020 Uttam Kumaran: pretty good. I know we didn’t make a ton of progress on transcripts.
146 00:12:51.160 ⇒ 00:12:55.979 Uttam Kumaran: you know. One thing that I’m finding is like, you know, the more products that we release.
147 00:12:56.200 ⇒ 00:13:01.610 Uttam Kumaran: And the more features we release, like we have to bake in some time for like
148 00:13:01.720 ⇒ 00:13:10.550 Uttam Kumaran: fixes and leveling up right like the migration, for example, that we’re planning is not something that like was on our immediate roadmap.
149 00:13:10.930 ⇒ 00:13:15.049 Uttam Kumaran: And so one thing I think we have to try to think about as a team is like.
150 00:13:15.240 ⇒ 00:13:22.370 Uttam Kumaran: how can we balance new stuff versus, you know, maintaining existing?
151 00:13:22.760 ⇒ 00:13:32.680 Uttam Kumaran: I think you know, originally we were like probably 80% new, 90% new. Now, we’re kind of closer, probably to 30 to 50% new.
152 00:13:32.810 ⇒ 00:13:38.440 Uttam Kumaran: And the rest of the time is going to like fixing the system. And then, of course, like this team is unique. Because
153 00:13:39.076 ⇒ 00:13:45.170 Uttam Kumaran: I mean, I I don’t say unique. But with we will have kind of a similar motion on the data side. But
154 00:13:45.460 ⇒ 00:14:02.920 Uttam Kumaran: the time is split between work for clients and work for us internally so, and like the work internal, is always going to be second for whatever the immediate client thing is. So I think that’s also helpful just to note and maybe, Sam, when we talk we can kind of take a look at like.
155 00:14:03.110 ⇒ 00:14:09.460 Uttam Kumaran: okay, what is our, what is our true capacity? And how can we make sure that we
156 00:14:09.770 ⇒ 00:14:11.710 Uttam Kumaran: we take on the right amount of work?
157 00:14:13.432 ⇒ 00:14:15.990 Uttam Kumaran: And keep space. So yeah.
158 00:14:18.040 ⇒ 00:14:35.539 Sam Roberts: Yeah, I I definitely am. I’m feeling that a little. I mean, you know, it’s been 2 weeks. So I’m still kind of like I I have a much better sense of like the company and everything now. But I’m definitely feeling like, Yeah, there’s, you know, some of the things that I was seeing. That’s kind of why I was thinking with this migration to next would make it a little bit easier to keep moving this forward rather than like.
159 00:14:36.180 ⇒ 00:14:50.909 Sam Roberts: Oh, well, I’m gonna have to jump into that and do this and like, get that back end going and get that front end like this will streamline that a little bit. So it’ll kind of take things off that. But yeah, it definitely is. You know, the secondary nature of it is kind of that, you know. There’s always things that are
160 00:14:51.380 ⇒ 00:14:55.820 Sam Roberts: what’s the word here nice to have and and quality
161 00:14:56.140 ⇒ 00:15:03.880 Sam Roberts: improvements? And so like a lot of the like internal stuff is basically that that the client stuff does. I get you exactly what you’re saying. That’s where the
162 00:15:04.100 ⇒ 00:15:11.550 Sam Roberts: that’s where the money comes from. That’s what you gotta focus on but in like, at any company, you know, even a product company, there’s always like technical debt and things that are like.
163 00:15:13.060 ⇒ 00:15:25.880 Sam Roberts: When when do you prioritize it? And oftentimes it’s it’s later than it should be for that reason. But it’s just because the nature of. But I think I’m getting a better sense of like where someone think the lifts are that it’ll actually help
164 00:15:28.410 ⇒ 00:15:30.930 Sam Roberts: keep that balance in a good way.
165 00:15:35.200 ⇒ 00:15:52.066 Sam Roberts: yeah. Trying to think, you know, cause even even just like the the co-pilot Kit test was a little bit like just getting the stuff in there and like, now, I’m seeing like, okay, this can go here. But you know, those agents are good, but with the na, then it’s different. And just understanding how some of these tools put together, and even just saying, like in the future, how things will go like we were just talking about that
166 00:15:53.190 ⇒ 00:15:54.980 Sam Roberts: the timeout issue.
167 00:15:55.330 ⇒ 00:15:55.730 Uttam Kumaran: Yeah.
168 00:15:56.270 ⇒ 00:16:08.500 Sam Roberts: That that is, I don’t know. If do we have a lot of long running, and it end things that we’re gonna have to be like testing like this. You know, there’s just little things like that that. I’m like, oh, we wouldn’t have known that. But now we’re gonna know that moving forward. How do we maintain that kind of like.
169 00:16:08.770 ⇒ 00:16:09.290 Uttam Kumaran: Yeah.
170 00:16:09.290 ⇒ 00:16:14.209 Sam Roberts: Awareness of of, you know things that will be.
171 00:16:14.970 ⇒ 00:16:16.419 Sam Roberts: what am I trying to say here, like
172 00:16:17.440 ⇒ 00:16:19.540 Sam Roberts: we’re gonna build a lot of in and in flows.
173 00:16:19.980 ⇒ 00:16:28.120 Sam Roberts: Some of them are going to be longer than 100 seconds which sometimes doesn’t matter but like in this case, like, are we gonna be wanting to test them more often.
174 00:16:28.330 ⇒ 00:16:30.350 Sam Roberts: you know, for other clients and things.
175 00:16:30.790 ⇒ 00:16:33.670 Sam Roberts: Do we have to be aware it’s just like things like that that are like
176 00:16:34.060 ⇒ 00:16:39.540 Sam Roberts: a little bit of the like. Step back and think of the process of our building processes. Kind of thing.
177 00:16:39.540 ⇒ 00:16:39.900 Uttam Kumaran: Yeah.
178 00:16:39.900 ⇒ 00:16:40.440 Sam Roberts: Oh.
179 00:16:41.270 ⇒ 00:16:45.743 Sam Roberts: yeah, sorry. I’m just kind of rambling at this point, because I’m on. My brain is kind of going in a hundred directions. But
180 00:16:45.930 ⇒ 00:16:51.030 Uttam Kumaran: I also think there’s like some flows that certainly we can graduate out of Nan
181 00:16:51.210 ⇒ 00:17:08.400 Uttam Kumaran: like as thing as things mature like. I feel like, you know. Ultimately, a lot of like these are all reproducible in code. I think probably 2 kind of core things is one, that the speed at which we can prototype
182 00:17:08.569 ⇒ 00:17:17.469 Uttam Kumaran: and test, and in an end is really high. So kind of like. The only way we can compete with that is like, if we have a developer lifecycle
183 00:17:17.710 ⇒ 00:17:38.189 Uttam Kumaran: or a developer sort of set of tooling and framework that allows people to do rapid prototyping, but also do this purely in code, which tough, I think the second piece, though, is that we should maybe instead have a path towards hey things that are stable. Maybe they should move into a
184 00:17:38.410 ⇒ 00:17:45.319 Uttam Kumaran: like whatever. However, we’re gonna write our agents and code, you know, that may
185 00:17:45.760 ⇒ 00:17:52.839 Uttam Kumaran: give us more observability. That may allow us to test changes.
186 00:17:55.700 ⇒ 00:18:03.660 Uttam Kumaran: you know, and and I don’t know. I’m sure there’s a slew of improvements there, I think probably the biggest, the biggest thing we’re gonna have is, I think.
187 00:18:04.200 ⇒ 00:18:09.889 Uttam Kumaran: like. For for me, my technical skill set is writing SQL. Python.
188 00:18:10.300 ⇒ 00:18:13.669 Uttam Kumaran: So what? And I think, even for the team, I think
189 00:18:14.290 ⇒ 00:18:19.089 Uttam Kumaran: both Casey and the software have now figured out like the ways to by code everything. But
190 00:18:19.230 ⇒ 00:18:23.869 Uttam Kumaran: probably what we’ll have to think about is like, how can we create a set of like
191 00:18:24.429 ⇒ 00:18:40.290 Uttam Kumaran: AI powered like tools. So anyone can write to that repo, for example, like one of the things I’m interested in is as soon as the sort of mono repo, or whatever is figured out like, I want to start trying to ship features my myself, and sort of understand like how to
192 00:18:40.430 ⇒ 00:18:46.240 Uttam Kumaran: test things locally, end to end and and start to use the agents to develop features basically end to end
193 00:18:46.967 ⇒ 00:18:55.629 Uttam Kumaran: sort of push that because then I’ll be able to see like what’s possible. But that is something that like, let’s say, we on board a developer and
194 00:18:56.160 ⇒ 00:19:04.789 Uttam Kumaran: like also our team, and what I kind of want it to be like. Whether or whether or not they know how to write great code. It doesn’t matter
195 00:19:05.170 ⇒ 00:19:12.169 Uttam Kumaran: like our the way we do things will will level. That will sort of level the playing field, you know.
196 00:19:13.050 ⇒ 00:19:17.550 Sam Roberts: Yeah, I I mean, I’ve had a lot of thoughts about even just jumping into this stuff myself and thinking, like.
197 00:19:17.660 ⇒ 00:19:45.951 Sam Roberts: you know, just some like readmes for how to get things running, because, like I, I’m able to figure it out because I have done enough web development to know where to look kind of thing. But if someone is newer to stuff, or it’s coming from a python, you know, there’s like there’s a lot of things that are like could be laid out. And, like even just, I just had cursor generate a couple of readmes for the different repos, and that was almost enough. But I think between that a bit of memory banking stuff, so that people know like what’s in progress, what’s still to do things like that?
198 00:19:46.830 ⇒ 00:19:54.440 Sam Roberts: adding, you mentioned like a linear Mcp. I was looking at that a little bit because that will help. I think I haven’t played enough with like Mcp tools and cursor to know.
199 00:19:54.770 ⇒ 00:20:07.240 Sam Roberts: like what is helpful and what is just, you know, gonna confuse the the Om, because it has too many tools or something. But the other thing is just cursor rules per project per
200 00:20:07.692 ⇒ 00:20:12.059 Sam Roberts: per repo, I think will help a lot, because I have. I have a few like
201 00:20:12.880 ⇒ 00:20:21.930 Sam Roberts: custom ones that I have set up that are maybe helping me a little bit in some of these projects, but once once everything is kind of standardized. It’ll be
202 00:20:22.400 ⇒ 00:20:31.040 Sam Roberts: nicer to say like, Oh, this is where this lives. This is where again, cursor, you know, can know that. And then people should be able to jump in a bit more easily, because that’ll
203 00:20:31.820 ⇒ 00:20:34.939 Sam Roberts: alongside the code. And then, as long as you’re opening cursor
204 00:20:35.470 ⇒ 00:20:40.570 Sam Roberts: to that repo, it knows what to do. So I definitely like trying to think about.
205 00:20:40.770 ⇒ 00:20:44.530 Sam Roberts: yeah, like, developer productivity kind of stuff.
206 00:20:45.140 ⇒ 00:20:49.369 Sam Roberts: you know, AI assisted, but specifically like with the tooling that we have.
207 00:20:49.740 ⇒ 00:20:50.160 Uttam Kumaran: Yeah.
208 00:20:50.160 ⇒ 00:20:54.482 Sam Roberts: Been like top of mind for me, because I’ve been doing it right now and getting up and running
209 00:20:55.070 ⇒ 00:20:59.830 Sam Roberts: so I I think there’s a lot of room for improvement there. I’m already like
210 00:21:00.150 ⇒ 00:21:02.807 Sam Roberts: thinking about the stuff as I’m doing it.
211 00:21:04.820 ⇒ 00:21:20.300 Sam Roberts: And that would have been part of my like put plan for a migration. But now that I’m getting the migration altogether almost in Gpd. 5, I’m thinking I’ll just start adding that stuff to that repo like, get some cursor rules in there. Get some. Get the read me going. And it’ll be a lot easier for someone like yourself to jump in
212 00:21:20.620 ⇒ 00:21:26.287 Sam Roberts: prototype something you know. Add, add a feature. Do something
213 00:21:27.110 ⇒ 00:21:29.300 Sam Roberts: pull in more data, whatever it is.
214 00:21:29.540 ⇒ 00:21:33.340 Sam Roberts: once it’s all together, and it’s 1 place to go.
215 00:21:34.650 ⇒ 00:21:38.530 Sam Roberts: So I’m I’m pretty excited about this. I’m pretty excited about, you know.
216 00:21:38.960 ⇒ 00:21:45.850 Sam Roberts: now that it’s all together adding the cursor rules, adding some, maybe some Mcp stuff, because I was even trying to figure out how to get it to
217 00:21:46.260 ⇒ 00:21:58.189 Sam Roberts: automate some of the testing, you know, especially as I’m migrating this thing over. I wanted it to look exactly the same. And I’m like, why can’t cursor know that like this? And I found like a browser? Mcp? There’s a lot of different stuff that we can start looking into. That might make this
218 00:21:59.160 ⇒ 00:21:59.940 Sam Roberts: much.
219 00:22:00.530 ⇒ 00:22:02.640 Sam Roberts: I won’t say simpler, but like
220 00:22:03.650 ⇒ 00:22:08.569 Sam Roberts: user friendly for someone that might not know exactly how the whole system works. Already.
221 00:22:08.830 ⇒ 00:22:12.368 Uttam Kumaran: Yeah, exactly. I just think one thing that we’re gonna have to think about is like.
222 00:22:13.250 ⇒ 00:22:18.410 Uttam Kumaran: and it’s something I’m thinking on. A lot of teams like, should we pause other stuff and focus on training
223 00:22:18.750 ⇒ 00:22:19.560 Uttam Kumaran: for a sec.
224 00:22:19.560 ⇒ 00:22:20.210 Sam Roberts: Hmm.
225 00:22:20.500 ⇒ 00:22:24.740 Uttam Kumaran: Or should we keep building new features like
226 00:22:25.880 ⇒ 00:22:36.920 Uttam Kumaran: right? Should we? Should we, instead be like, okay, well, we’re gonna pause. New features. We’re gonna level everyone up on like, how to use cursor for everything like just in even in just in our team.
227 00:22:37.430 ⇒ 00:22:37.820 Sam Roberts: Yeah.
228 00:22:37.820 ⇒ 00:22:42.430 Uttam Kumaran: Do a sprint on like, okay, we want to create an environment that
229 00:22:42.750 ⇒ 00:22:49.180 Uttam Kumaran: the barrier between writing and pushing code is like 20 to 50% less.
230 00:22:49.570 ⇒ 00:22:50.570 Sam Roberts: Yeah, yeah.
231 00:22:51.140 ⇒ 00:22:57.510 Uttam Kumaran: And we’ve put a pause on like new stuff for a sec while we figure that out, because ideally, that should accelerate
232 00:22:57.910 ⇒ 00:22:59.789 Uttam Kumaran: any of the new stuff right?
233 00:23:00.590 ⇒ 00:23:01.300 Sam Roberts: Yeah.
234 00:23:02.170 ⇒ 00:23:09.559 Uttam Kumaran: For example, if you told me like, Hey I there’s a there’s a clear path that I can even start to push stuff.
235 00:23:10.020 ⇒ 00:23:15.199 Uttam Kumaran: Yeah, that’d be great, right, and and other people on the other people internally.
236 00:23:15.430 ⇒ 00:23:18.829 Uttam Kumaran: as they want to add features, they can add those, you know.
237 00:23:19.510 ⇒ 00:23:20.110 Sam Roberts: Yeah.
238 00:23:20.940 ⇒ 00:23:35.330 Sam Roberts: yeah, I I kinda I get what you’re saying that like push and pull between, like, you know, I I was always talking about when I was like working on other stuff was like to get the quote for what it was like. You know, if I had time to chop down a tree a certain amount of time talking a tree. I spend 90% of it sharpening the axe. Kind of thing.
239 00:23:35.570 ⇒ 00:23:36.570 Uttam Kumaran: Yeah.
240 00:23:36.740 ⇒ 00:23:43.590 Sam Roberts: And I’m like that is, you know, that pays dividends, and especially when we’re setting these things up. And we need to know how to use the tooling.
241 00:23:43.900 ⇒ 00:23:49.919 Sam Roberts: I think there’s something to be said for that. But I think there’s also especially like
242 00:23:50.350 ⇒ 00:24:04.450 Sam Roberts: with us here on this team, like we. We kind of know how to use some of this stuff. We just have to get the things in place to make it easier, not only to for us to use them, but to onboard someone and bring someone in and
243 00:24:04.630 ⇒ 00:24:16.588 Sam Roberts: get them up and running rather than like, you know. Let’s hit the pause. Let’s all get up, you know, level up a little bit, but I think like we can be leveling up as we’re going, and then by doing
244 00:24:17.890 ⇒ 00:24:19.710 Sam Roberts: not documentation, but just like
245 00:24:19.870 ⇒ 00:24:43.690 Sam Roberts: adding these features, adding the cursor rules, adding the readmes, things like that are gonna gonna help anyone getting into the repo that are also gonna help us working with it will help new people coming on, and then they’ll know how to, you know. Maybe there’s a little bit of a onboarding of like. Oh, you haven’t used cursor! Well, here’s how it’s set up. Here’s how you have to have, you know, but like for us, like we kind of already have that. I think it’s just adding the right pieces in for us to keep moving.
246 00:24:43.960 ⇒ 00:24:44.500 Sam Roberts: Oh.
247 00:24:44.500 ⇒ 00:24:45.230 Uttam Kumaran: Yes.
248 00:24:46.480 ⇒ 00:25:02.999 Sam Roberts: So I don’t know if it’s necessarily like for us, at least, for you know at least speaking for for this team here. I don’t know how other people I was a little bit curious about who would use cursor and stuff on the engineering call the other day, because I was curious, like how other people are using it. That’s a whole other thing. I’m not. I don’t have as much insight into, obviously
249 00:25:03.960 ⇒ 00:25:07.950 Sam Roberts: for other projects and for the data side. I’m much blinder there, but.
250 00:25:08.440 ⇒ 00:25:17.459 Uttam Kumaran: Yeah, for the data side. This is where, like, I think, between me, you and a wish, we’ll bridge the gap a little bit. But that’s that is our largest
251 00:25:17.580 ⇒ 00:25:21.960 Uttam Kumaran: sort of cost center is time spent on on data. So
252 00:25:22.110 ⇒ 00:25:42.836 Uttam Kumaran: one thing that, like one thing that we that this team, you know, hopefully, can start to own is like, how does any engineer Brainforge just become, I mean, how does any employee get any AI enabled here? And then how does every engineer? How do we make sure that for whatever their engineering task is, there is a way to augment with AI
253 00:25:44.050 ⇒ 00:25:45.020 Uttam Kumaran: you know.
254 00:25:47.421 ⇒ 00:25:56.090 Uttam Kumaran: So I think overall, it may just be like, look until we can get like another one or 2 engineers working on AI stuff.
255 00:25:56.300 ⇒ 00:25:59.279 Uttam Kumaran: Maybe we we just have to go a little bit slower
256 00:25:59.620 ⇒ 00:26:05.670 Uttam Kumaran: and like, I mean. But I do think that if we can start to enable ourselves to ship
257 00:26:06.970 ⇒ 00:26:10.080 Uttam Kumaran: 25, or 50% faster than.
258 00:26:10.080 ⇒ 00:26:10.680 Sam Roberts: Yeah.
259 00:26:11.060 ⇒ 00:26:24.509 Uttam Kumaran: Maybe we don’t. Maybe we can push out the like need for the next person, or solve our own problem like, put some constraints on, you know, because I don’t know if we’ll be able to bring on another AI person until another 2 clients or so sign
260 00:26:24.790 ⇒ 00:26:25.199 Sam Roberts: Sure, sure.
261 00:26:25.581 ⇒ 00:26:33.600 Uttam Kumaran: So I’m kind of like, okay, maybe we just use that as constraint. We focus on improving our own development speeds
262 00:26:34.468 ⇒ 00:26:41.463 Uttam Kumaran: and that as a result, should allow.
263 00:26:43.470 ⇒ 00:26:46.580 Uttam Kumaran: you know, should at least at least push that a little bit further down the line.
264 00:26:47.910 ⇒ 00:26:49.180 Sam Roberts: Yeah, I think that’s
265 00:26:49.310 ⇒ 00:26:53.629 Sam Roberts: probably the right way to think about it. I mean, concerns are powerful that way. But also, you know, like.
266 00:26:54.330 ⇒ 00:26:59.630 Sam Roberts: you know, I I think it also depends on maybe, like you and I, and and a ways can really talk about like where I can support
267 00:26:59.760 ⇒ 00:27:02.680 Sam Roberts: the supporting of things. If that makes sense like.
268 00:27:02.680 ⇒ 00:27:03.450 Uttam Kumaran: Yes.
269 00:27:03.650 ⇒ 00:27:10.889 Sam Roberts: Like, where like is it? You know? Where’s my time best spent getting, you know, helping lift everyone up a little bit
270 00:27:11.000 ⇒ 00:27:29.880 Sam Roberts: like maybe maybe getting into like chatting with someone on the data side and just getting a sense of how they’re using the tools and and what they’re not using. And you know, I I that’s you know, it’s just I don’t have a great sense of. And so I think as I get more of that, I can start being a little more helpful broadly but also like
271 00:27:30.220 ⇒ 00:27:34.339 Sam Roberts: leveling us up with with, you know, cursor rules with with
272 00:27:34.630 ⇒ 00:27:39.789 Sam Roberts: standardized. You know, this is how we’re just gonna do when we need to launch this kind of thing, we’re gonna focus on
273 00:27:39.830 ⇒ 00:28:06.830 Sam Roberts: these tools. Or, you know, like like you were saying, this is that something I was very curious about was like, how much do we want to lean on any end? I think hearing you say like once. It’s kind of like matured stable moving. That to code is something I was curious about, especially as I was digging into co-pilot kit and seeing the other frameworks that it that it hooks into. You know there’s there’s other other things out there, and I wasn’t sure where we wanted to go with it just as a company as a where that is. So I think that’s stuff I can get a better sense of now.
274 00:28:07.430 ⇒ 00:28:08.300 Uttam Kumaran: Yeah, okay.
275 00:28:08.680 ⇒ 00:28:20.350 Sam Roberts: Like Stan, like standard like, Oh, we’re launching this kind of project. We will be using this tooling. We’ll be using these frameworks. We have the cursor rules in this repo to start a new project. That that sort of stuff, I think, is, gonna be
276 00:28:21.730 ⇒ 00:28:25.939 Sam Roberts: very helpful. And I I don’t have a great sense for all that. Yet on the client side or the data.
277 00:28:26.390 ⇒ 00:28:28.700 Uttam Kumaran: But I think I can definitely relate.
278 00:28:28.700 ⇒ 00:28:29.310 Sam Roberts: What’s on that?
279 00:28:29.540 ⇒ 00:28:34.990 Uttam Kumaran: I think longer term. Probably most of our time will go into planning.
280 00:28:35.270 ⇒ 00:28:37.810 Uttam Kumaran: Yeah. And then we’ll have, like the agent.
281 00:28:38.130 ⇒ 00:28:43.720 Uttam Kumaran: probably like one of the things that cursor releases like they’re they’re
282 00:28:44.250 ⇒ 00:28:56.310 Uttam Kumaran: basically their cursor agent. But like, yeah, the more time we spend on planning or we have AI help us plan. We’ll then have AI actually take the 1st stab at like the entire Pr.
283 00:28:56.970 ⇒ 00:29:07.409 Uttam Kumaran: right? And and that’s the stuff I think, like it’s gonna be huge right. If if we find out that for any new feature we develop, let’s spend an hour.
284 00:29:07.950 ⇒ 00:29:11.650 Uttam Kumaran: or to write a great like prd.
285 00:29:11.920 ⇒ 00:29:20.429 Uttam Kumaran: and then have AI take a stab at like the technical requirements to fulfill it that gets approved, and then AI takes both of those and then
286 00:29:20.550 ⇒ 00:29:22.699 Uttam Kumaran: use cloud code to like, write the pr
287 00:29:24.880 ⇒ 00:29:25.490 Sam Roberts: Yeah.
288 00:29:25.850 ⇒ 00:29:29.149 Uttam Kumaran: Right like that takes probably something that would have taken
289 00:29:30.200 ⇒ 00:29:35.320 Uttam Kumaran: so 5 to 10 h, 3 days into like maybe half day.
290 00:29:35.870 ⇒ 00:29:38.100 Sam Roberts: Yeah, no, I think that’s exactly exactly right.
291 00:29:38.100 ⇒ 00:29:41.910 Uttam Kumaran: Our crew has to push the limits right? So this is where it’s like, it’s not just
292 00:29:42.170 ⇒ 00:29:46.749 Uttam Kumaran: can we? It’s not just can we
293 00:29:47.750 ⇒ 00:29:50.250 Uttam Kumaran: like, just use cursor in the sidebar.
294 00:29:50.440 ⇒ 00:30:00.980 Uttam Kumaran: I’m thinking, like, okay, can we have, like an agent that actually can take on like simple tasks? Right? We start off with like, Okay, are these one or 2 point tasks? Those should all go to AI,
295 00:30:01.220 ⇒ 00:30:02.800 Uttam Kumaran: and and we should.
296 00:30:03.340 ⇒ 00:30:09.480 Uttam Kumaran: We should all try to use like, have the agent basically build it end to end and see what happens.
297 00:30:10.168 ⇒ 00:30:15.190 Uttam Kumaran: Because I have a feeling. Everybody on every team is. Gonna figure out how important cut cursor is.
298 00:30:16.015 ⇒ 00:30:19.480 Uttam Kumaran: Like that is not? That is like table stakes.
299 00:30:20.250 ⇒ 00:30:22.329 Sam Roberts: Of course, of course. Yeah, yeah.
300 00:30:22.650 ⇒ 00:30:23.089 Uttam Kumaran: Like we.
301 00:30:23.090 ⇒ 00:30:23.900 Sam Roberts: Yeah, I mean, I think.
302 00:30:23.900 ⇒ 00:30:28.069 Uttam Kumaran: But the best place they’re saying, which is like, have a planning agent, have, like
303 00:30:28.320 ⇒ 00:30:31.879 Uttam Kumaran: blog code, then execute maybe, or or whatever you know.
304 00:30:32.690 ⇒ 00:30:42.989 Sam Roberts: Yeah, no, you’re absolutely right. I’m definitely thinking, like medium term, more than longer term as I’m talking here. But I I see what you’re saying. I think a lot of that comes also with just like.
305 00:30:43.820 ⇒ 00:30:48.160 Sam Roberts: I mean, that’s just good engineering practices like planning. You know what I mean like that’s
306 00:30:48.510 ⇒ 00:30:52.959 Sam Roberts: huge. And then, like what runs it, whether it’s a developer or an AI agent is like
307 00:30:54.150 ⇒ 00:31:05.580 Sam Roberts: just another piece you can plug into that. But like the pre planning process. The pre planning and planning process is, you know, a huge chunk. I think, that can go into building good bid systems in general.
308 00:31:05.980 ⇒ 00:31:06.380 Uttam Kumaran: Yeah, so.
309 00:31:06.380 ⇒ 00:31:09.970 Sam Roberts: I’m I’m with you on that, I think, setting us up to do that.
310 00:31:10.140 ⇒ 00:31:15.649 Sam Roberts: You know, you want to create a good engineering environment where like, that’s part of the process, to begin with, anyway.
311 00:31:16.665 ⇒ 00:31:20.480 Sam Roberts: And then, as as we plug into
312 00:31:20.820 ⇒ 00:31:31.369 Sam Roberts: more agents, or even planning tools, and whatever it is whatever agents we’re having to help us write that or look at what we’ve already done. And you know, scaffold things out based on that, like, I think that’s
313 00:31:32.390 ⇒ 00:31:33.790 Sam Roberts: that that that
314 00:31:33.920 ⇒ 00:31:40.009 Sam Roberts: dovetails very well together. As long as the you know, engineering culture is already there for a planning culture and a
315 00:31:40.130 ⇒ 00:31:43.409 Sam Roberts: you know, which I think we’re. We’re getting good at.
316 00:31:43.410 ⇒ 00:31:45.579 Uttam Kumaran: We’re we’re bet we’re better than most.
317 00:31:46.010 ⇒ 00:31:46.580 Sam Roberts: Yeah.
318 00:31:46.710 ⇒ 00:31:48.849 Uttam Kumaran: But that bar is very low, like we.
319 00:31:49.119 ⇒ 00:32:00.970 Sam Roberts: Exactly. I mean, that’s like coming from like the mechanical engineering background I had like, that was like the biggest thing in software. It’s like there are, you know, when you’re doing mechanical engineering. And there’s like physical things. And it’s like, you know.
320 00:32:00.990 ⇒ 00:32:01.710 Uttam Kumaran: The, yeah.
321 00:32:01.710 ⇒ 00:32:08.560 Sam Roberts: Of constraints. You have to do that. There’s no other way to do it, really but when it comes to software like, you have a lot of room to like
322 00:32:08.820 ⇒ 00:32:11.709 Sam Roberts: scoot under that. And plenty of people do.
323 00:32:12.170 ⇒ 00:32:14.449 Sam Roberts: Yeah works for a lot of things. But when you want.
324 00:32:14.450 ⇒ 00:32:19.380 Uttam Kumaran: But see like for me, this is where, like, I’m gonna in the
325 00:32:20.370 ⇒ 00:32:26.920 Uttam Kumaran: all the characters that we’re playing in this movie for me. My job is to keep the foot on the gas
326 00:32:27.699 ⇒ 00:32:33.389 Uttam Kumaran: and like I say that as I can look at myself and say that that is like
327 00:32:33.880 ⇒ 00:32:35.839 Uttam Kumaran: I have to do that actively.
328 00:32:35.950 ⇒ 00:32:38.049 Uttam Kumaran: And I kind of do that.
329 00:32:38.620 ⇒ 00:33:06.590 Uttam Kumaran: just like we’re kind of robotically, because that is my job in the company right like that is what I’m as a as an employee. My job is to keep the heartbeat, you know, and so it’s tough, because I was on the other side where? And I see the stuff we build like it’s it’s not half baked because our quality was bad. It’s that we’re building a hundred things at the same time with just a couple of people part kind of part time, because we’re working on client work. Right? So
330 00:33:06.710 ⇒ 00:33:11.710 Uttam Kumaran: it’s it’s tough, and we are constrained. And so for me.
331 00:33:11.820 ⇒ 00:33:25.019 Uttam Kumaran: it’s like, okay, we have those constraints. Let’s see the limit we can get with AI, and then ultimately look, all those things are gonna help us develop for clients faster. And then the the faster we develop our clients. Our bet is that
332 00:33:25.170 ⇒ 00:33:27.809 Uttam Kumaran: more clients come through the door. Their margins are higher.
333 00:33:28.380 ⇒ 00:33:29.230 Uttam Kumaran: So
334 00:33:29.540 ⇒ 00:33:34.959 Uttam Kumaran: I I don’t. I don’t know Casey or Masa like, what do you guys think? I know we just are at time, but
335 00:33:35.240 ⇒ 00:33:38.489 Uttam Kumaran: maybe we should spend next sprint sort of focused on that.
336 00:33:41.650 ⇒ 00:33:48.300 Casie Aviles: Yeah, I do think so as well. I mean, I I agree with the piece that you’re mentioning, that you know we’re kind of.
337 00:33:48.710 ⇒ 00:33:51.690 Casie Aviles: you know, split between 2 things. So
338 00:33:51.980 ⇒ 00:34:02.226 Casie Aviles: like currently right now that I’m working on this task and I didn’t expect to take longer than it than it should. And that’s primarily because, you know,
339 00:34:02.950 ⇒ 00:34:08.120 Casie Aviles: there’s like a phase where we just kept pushing to group like
340 00:34:08.389 ⇒ 00:34:10.689 Casie Aviles: produce more, you know, new stuff.
341 00:34:11.270 ⇒ 00:34:17.199 Casie Aviles: But we never really. And you know, we started noticing that a lot of technical debt has been piling up.
342 00:34:17.690 ⇒ 00:34:21.470 Casie Aviles: and that’s kind of what’s frustrating with with the whole
343 00:34:22.125 ⇒ 00:34:27.160 Casie Aviles: client hub part as well. There was a lot of technical debt, and you know.
344 00:34:27.730 ⇒ 00:34:29.940 Casie Aviles: like, yeah. And I also agree that
345 00:34:30.710 ⇒ 00:34:35.929 Casie Aviles: with what Sam said that there would be a there would have to be a lot of planning first, st as well.
346 00:34:39.100 ⇒ 00:34:41.410 Mustafa Raja: Yeah, I like this suggestion, too.
347 00:34:44.040 ⇒ 00:34:44.639 Uttam Kumaran: Yeah.
348 00:34:44.800 ⇒ 00:34:47.769 Uttam Kumaran: Okay, so let me think a little bit about it. This weekend
349 00:34:48.631 ⇒ 00:34:53.080 Uttam Kumaran: maybe we can. I’ll I’ll make some hard decisions on what we’re gonna put a pause on.
350 00:34:53.370 ⇒ 00:34:59.179 Uttam Kumaran: I still think, like, probably my, only ask the co-pilot kid stuff it would actually make
351 00:34:59.660 ⇒ 00:35:12.989 Uttam Kumaran: like some of those agents usable. So if there, if there is like a even simple plug and play way, or you can be like, Hey, it’s set up. You go make sure the prompts and stuff are working in the cloud. I can handle that
352 00:35:14.660 ⇒ 00:35:15.390 Uttam Kumaran: but
353 00:35:15.740 ⇒ 00:35:21.539 Uttam Kumaran: and then I can look at like what else is remaining, and maybe plan out some tickets where we can just focus on
354 00:35:21.690 ⇒ 00:35:27.069 Uttam Kumaran: the developer lifecycle. So again, those will include if we were start
355 00:35:27.610 ⇒ 00:35:33.409 Uttam Kumaran: from the earliest, it’d be AI for planning AI for technical design.
356 00:35:35.700 ⇒ 00:35:42.219 Uttam Kumaran: Some simple agents that I can actually execute and create Prs AI for Pr reviews.
357 00:35:42.440 ⇒ 00:35:43.250 Uttam Kumaran: Yeah.
358 00:35:43.840 ⇒ 00:35:47.490 Uttam Kumaran: And then I don’t know if there is an AI, for like
359 00:35:48.370 ⇒ 00:35:58.430 Uttam Kumaran: like visual, like, for example, like someone, can AI even check out the demo site and be like, this is what we wanted. I don’t know what I have. We would have to research like, what are the
360 00:36:00.010 ⇒ 00:36:05.459 Uttam Kumaran: what are the options right now for front end like AI sort of testing.
361 00:36:06.000 ⇒ 00:36:09.620 Sam Roberts: Yeah, I was. I was actually digging into that a little bit yesterday because I was trying to figure out if there’s a way to plug that in
362 00:36:09.620 ⇒ 00:36:10.239 Sam Roberts: like we couldn’t.
363 00:36:10.240 ⇒ 00:36:11.549 Sam Roberts: As I was like getting this.
364 00:36:12.010 ⇒ 00:36:16.310 Uttam Kumaran: We could push it on to the design team to Qa.
365 00:36:16.550 ⇒ 00:36:17.510 Uttam Kumaran: But.
366 00:36:17.870 ⇒ 00:36:19.460 Sam Roberts: Yeah, but still.
367 00:36:19.460 ⇒ 00:36:24.000 Uttam Kumaran: The thing is like, not a lot of our stuff is coming from the design team in terms of
368 00:36:24.270 ⇒ 00:36:28.230 Uttam Kumaran: like cause. We just sort of move quickly on stuff. But
369 00:36:30.130 ⇒ 00:36:35.039 Uttam Kumaran: yeah, I think we’ll have to think about. But again, if we hit 4 out of 5 of those like, I feel pretty good, you know.
370 00:36:35.450 ⇒ 00:36:37.689 Sam Roberts: Yeah, no, I think I think that’s the right
371 00:36:38.330 ⇒ 00:36:46.360 Sam Roberts: great way to think about it. I can definitely get that cop get stuff. Probably. I mean, if the agents are working like, I expect they’re working. I can probably plug that in today.
372 00:36:46.490 ⇒ 00:36:53.629 Sam Roberts: Okay, at least get it running on the copilotkit cloud, which has a
373 00:36:54.030 ⇒ 00:37:00.796 Sam Roberts: decent free tiers, so that we can at least test it. And then, if we need to move to the back end, that that might be a little bit more, but shouldn’t be much more.
374 00:37:03.040 ⇒ 00:37:06.894 Sam Roberts: But yeah, that that’s I mean, I think, yeah, think about that this weekend
375 00:37:07.860 ⇒ 00:37:11.290 Sam Roberts: Well, we we can convene Monday morning, I guess, like at our.
376 00:37:11.290 ⇒ 00:37:12.820 Uttam Kumaran: Yeah, plan, yeah.
377 00:37:12.820 ⇒ 00:37:19.630 Sam Roberts: Yeah, and then get a plan for the week from there. But I think this is, this is probably a good way to think about it like.
378 00:37:19.850 ⇒ 00:37:25.299 Sam Roberts: I don’t know if you know like this, I mean when you’re when you’re doing like software development, there’s like a ton of different methodologies. But like
379 00:37:25.420 ⇒ 00:37:32.009 Sam Roberts: one of them, that I’ve done a decent amount of times, not not necessarily in like a formalized way, but like a spike and stabilize kind of
380 00:37:33.410 ⇒ 00:37:35.529 Sam Roberts: way to go about it. And I think that’s kind of what we’re
381 00:37:35.730 ⇒ 00:37:43.791 Sam Roberts: getting here like we had a big spike in a lot of things. And now, it’s like, okay, there’s a lot of stuff out there. Let’s let’s get it stabilized.
382 00:37:44.060 ⇒ 00:37:44.590 Uttam Kumaran: Yes.
383 00:37:44.590 ⇒ 00:37:52.549 Sam Roberts: Which I think is is not a bad way to do it. But you gotta you gotta do the stabilization part. You can’t just be throwing everything out there all the time. So I think this is good.
384 00:37:56.410 ⇒ 00:37:57.620 Uttam Kumaran: Okay, perfect.
385 00:37:57.860 ⇒ 00:38:05.830 Uttam Kumaran: Alright? Well, thank you. Guys, yeah, I’ll I’ll follow up on slack with any open items. And then yeah, great conversation. Let’s plan for Monday, and then
386 00:38:05.950 ⇒ 00:38:10.820 Uttam Kumaran: if I don’t say Hi to everybody, well, I guess I’ll see everybody on a on a team call later. So
387 00:38:12.470 ⇒ 00:38:14.720 Uttam Kumaran: okay, alright, thank you.
388 00:38:14.720 ⇒ 00:38:16.220 Casie Aviles: Thank you. Thank you.
389 00:38:16.520 ⇒ 00:38:17.370 Sam Roberts: Everyone.
390 00:38:17.370 ⇒ 00:38:18.170 Uttam Kumaran: Bye.