Meeting Title: AI Platform Enablement Plan: Doc Review Date: 2025-10-07 Meeting participants: Casie Aviles, Uttam Kumaran, Mustafa Raja, Henry Zhao
WEBVTT
1 00:00:32.890 ⇒ 00:00:34.190 Uttam Kumaran: Hello!
2 00:00:35.790 ⇒ 00:00:37.549 Casie Aviles: Hey, how are your thoughts?
3 00:00:37.820 ⇒ 00:00:39.439 Uttam Kumaran: Hey dude, how are you?
4 00:00:40.140 ⇒ 00:00:41.330 Casie Aviles: Yeah, doing good.
5 00:00:41.910 ⇒ 00:00:42.589 Casie Aviles: How about you?
6 00:00:42.590 ⇒ 00:00:43.419 Uttam Kumaran: How’s the day?
7 00:00:44.490 ⇒ 00:00:45.370 Uttam Kumaran: Good.
8 00:00:46.300 ⇒ 00:00:49.609 Uttam Kumaran: Yeah. How’s it… how’s the day been going overall? Yeah, tell me.
9 00:00:50.360 ⇒ 00:00:53.440 Casie Aviles: Yeah, I think, you know, the…
10 00:00:53.970 ⇒ 00:01:03.879 Casie Aviles: workshop earlier with Robert went pretty well, although I didn’t… I wasn’t very active, but I was just, you know, absorbing as much as I can.
11 00:01:04.769 ⇒ 00:01:07.639 Uttam Kumaran: Nice. I’m jealous, dude, I gotta rewatch it.
12 00:01:07.640 ⇒ 00:01:08.890 Casie Aviles: Yeah. Yeah.
13 00:01:08.890 ⇒ 00:01:13.130 Uttam Kumaran: He’s really, really good at what he does, by the way, if you haven’t noticed.
14 00:01:13.490 ⇒ 00:01:16.510 Casie Aviles: Yeah… like…
15 00:01:17.760 ⇒ 00:01:23.600 Uttam Kumaran: I, I was never, like, a great analyst, but, I figured it out, but.
16 00:01:23.940 ⇒ 00:01:29.380 Uttam Kumaran: watching his type of work, I’m like, oh, okay, there’s, like, levels to this, like…
17 00:01:30.210 ⇒ 00:01:33.930 Casie Aviles: Yeah, definitely, and I like how it’s not, you know, like…
18 00:01:34.150 ⇒ 00:01:39.700 Casie Aviles: locking us into specific tools, and more about, you know, asking the big questions, I guess, and…
19 00:01:40.310 ⇒ 00:01:44.900 Uttam Kumaran: It’s all storytelling and, like, it’s a lot of creativity, actually.
20 00:01:44.900 ⇒ 00:01:45.760 Casie Aviles: Yeah.
21 00:01:46.480 ⇒ 00:01:47.250 Uttam Kumaran: Yeah, yeah.
22 00:01:48.000 ⇒ 00:01:48.980 Uttam Kumaran: Nice.
23 00:01:50.020 ⇒ 00:01:54.589 Henry Zhao: Yeah, I thought his README analysis was really good yesterday, that’s why I asked about it, I just wanted you all to see it.
24 00:01:55.750 ⇒ 00:01:57.230 Uttam Kumaran: Nice, okay, cool.
25 00:01:57.530 ⇒ 00:01:59.030 Henry Zhao: You’re best.
26 00:02:00.310 ⇒ 00:02:01.199 Uttam Kumaran: That’s great.
27 00:02:02.190 ⇒ 00:02:05.170 Henry Zhao: And because I did have a question on that funnel chart anyway, so that was legit.
28 00:02:07.650 ⇒ 00:02:27.399 Uttam Kumaran: Yeah, I’m… I just, like, sometimes I don’t know all the different, like, analysis methods. Like, I’m an engineer, so most of my analysis work that I did was, like, okay, like, there’s 5 ways of breaking this problem down, but I’m a… I’m… I used to do a lot of, like, financial analysis, so I think about drivers, I think about…
29 00:02:27.610 ⇒ 00:02:30.590 Uttam Kumaran: like, basically, I look at accelerations.
30 00:02:30.760 ⇒ 00:02:46.829 Uttam Kumaran: contribution to a different KPI, and that’s a lot of it. It’s sort of like taking a number and saying, okay, what are the 10 ways that this got affected? And you have to find the story. And there’s… but the thing is, you have to always assume there is a story. Like, for all of our clients, they’re…
31 00:02:47.110 ⇒ 00:02:51.509 Uttam Kumaran: they haven’t solved everything, and so that’s really what it starts with, I feel like.
32 00:02:55.630 ⇒ 00:02:58.250 Henry Zhao: I’m always worried about saying things that are, like, too obvious.
33 00:02:59.420 ⇒ 00:03:05.980 Uttam Kumaran: Well, I think the best way of doing it is, like, say… say the obvious thing up front to get everyone on the same page.
34 00:03:06.070 ⇒ 00:03:22.179 Uttam Kumaran: And then, like… and then all you have to ask is, like, okay, like, what are you… I just want to set the stage here, like, this is what we found. Does that… what I usually do is I basically, like, does that check out? Because, for example, when we went to default, right, we started to do some basic analysis, I was like, okay, it looks like you guys have, like.
35 00:03:22.180 ⇒ 00:03:34.920 Uttam Kumaran: this many customers. Does that check out? No. Okay, perfect, let’s start there. So let’s figure out, like, is our data wrong? Are our analysis wrong? Are they wrong? Because they don’t even know, right? So that’s a good way.
36 00:03:35.230 ⇒ 00:03:42.009 Uttam Kumaran: And, but for default, right, they’re so early for an insomnia, they might have tried everything, but no one will ever fault you for, like.
37 00:03:42.250 ⇒ 00:03:49.199 Uttam Kumaran: saying, okay, we’ve reached the same conclusions that y’all probably reached. Now it’s our job to, like, go one step further, you know?
38 00:03:49.670 ⇒ 00:03:50.190 Henry Zhao: Yeah.
39 00:03:51.540 ⇒ 00:03:57.669 Uttam Kumaran: The worst thing is if you spend, like, 2 months and arrive at the same thing that they already know, which is, like.
40 00:03:58.070 ⇒ 00:04:05.290 Uttam Kumaran: opens… email opens went down, and so raw revenue went down. It’s like, duh, like…
41 00:04:05.290 ⇒ 00:04:11.299 Henry Zhao: I mean, I think that’s a good point, and maybe before we start on a new client, to, like, align on what they already know, and what the gaps are.
42 00:04:11.530 ⇒ 00:04:16.609 Uttam Kumaran: But some of it is, like, the client is not gonna… not… it may not tell you that, like, up front. Yeah.
43 00:04:17.160 ⇒ 00:04:20.139 Henry Zhao: Sometimes it’s like, person number 3 knew, but you only met with person.
44 00:04:20.140 ⇒ 00:04:21.120 Uttam Kumaran: What’s important.
45 00:04:21.440 ⇒ 00:04:31.709 Uttam Kumaran: So the pro… often, dude, the problem is not that they can’t do this analysis, it’s just that they never get organized enough, and, like, nobody knows the whole story.
46 00:04:31.810 ⇒ 00:04:45.769 Uttam Kumaran: And us being really naive and coming in and being, like, asking super basic questions is actually, like, the solution. And nobody in their company can go after a problem with fresh eyes, like, other than us. Yeah.
47 00:04:45.780 ⇒ 00:04:53.640 Henry Zhao: And I think sometimes it also helps prioritize, it’s like, they know they have these 17 problems, but which ones are actually big issues, and which ones should we actually be tackling first?
48 00:04:54.140 ⇒ 00:04:58.759 Uttam Kumaran: That’s also the cases where sometimes they have, like, you have, like, a shiny object problem.
49 00:04:58.930 ⇒ 00:05:18.649 Uttam Kumaran: Where executive one… we’ve had many clients where executive wants this, but another executive wants this, and then you ask them… you work on it, and they’re like, why’d you spend this much time working on this? And you’re like, dude, why? You’re the one that told me to work on this. And so, we have that problem as well, so…
50 00:05:19.240 ⇒ 00:05:29.720 Uttam Kumaran: Yeah, that’s awesome, guys. I’m glad that, like, I think for Casey and for Mustafa, this area is a little bit new, so I’m really, really happy that you guys are getting exposure.
51 00:05:30.440 ⇒ 00:05:34.030 Uttam Kumaran: Cool, so I wanted just to…
52 00:05:34.200 ⇒ 00:05:41.779 Uttam Kumaran: First, I think I wanna… I’m gonna start doing more of these, which is kind of like, at a larger engineering organization.
53 00:05:41.860 ⇒ 00:05:45.419 Uttam Kumaran: There are, you know, common times where you do what’s called, like, a…
54 00:05:45.440 ⇒ 00:05:55.390 Uttam Kumaran: PRD review, or you basically review pay… you review writing, right? So, for example, I may come to the table with a proposal on how
55 00:05:55.390 ⇒ 00:06:09.919 Uttam Kumaran: we want to spend our time this quarter, and it… as someone who needs to propose the work, I have to write it down. And so usually we come to these meetings where people have… there’s a pre-read or something where you can read, and then it’s sort of like what’s called RFC, which is like Request for Comments.
56 00:06:09.920 ⇒ 00:06:21.369 Uttam Kumaran: And so, I want to start to get in the habit of this, you know, more on the engineering side, is where you… we take time to write down how we think the state of the world is, what our proposals are, and we kind of get comments.
57 00:06:21.370 ⇒ 00:06:36.240 Uttam Kumaran: I think similarly, we’ll do… we’ll do larger, like, architecture reviews as well, as we get more technical, and I get some more time to spend, like, on the solution architecture side. But I just want to kind of check if everybody was able to
58 00:06:36.300 ⇒ 00:06:43.370 Uttam Kumaran: to scan through the dock, or do we want to take 5 minutes, to just scan through the dock? It’s very, very…
59 00:06:43.650 ⇒ 00:06:51.789 Uttam Kumaran: simple. I think, Henry, I saw you… you did some in advance, but you’re probably the only one here that doesn’t have super familiarity with the platform, so that’s…
60 00:06:52.170 ⇒ 00:06:54.690 Uttam Kumaran: That was probably it, but Casey…
61 00:06:55.040 ⇒ 00:06:57.660 Uttam Kumaran: Mustafa, do you guys want to, like, take a sec to read?
62 00:06:59.390 ⇒ 00:07:00.260 Mustafa Raja: Oh, yes.
63 00:07:02.000 ⇒ 00:07:03.150 Casie Aviles: Can we just take a couple.
64 00:07:03.150 ⇒ 00:07:05.949 Uttam Kumaran: minutes, yeah. And then maybe in the…
65 00:07:06.090 ⇒ 00:07:15.179 Uttam Kumaran: In the meantime, Henry, do you have any questions about, like, the history of the platform and things like that that I can answer in the meantime while those guys sort of take a…
66 00:07:15.180 ⇒ 00:07:17.840 Henry Zhao: I’m sure we’ll get the context.
67 00:07:17.840 ⇒ 00:07:18.230 Uttam Kumaran: Okay.
68 00:07:18.230 ⇒ 00:07:18.890 Henry Zhao: Got it.
69 00:07:19.330 ⇒ 00:07:38.649 Uttam Kumaran: So basically, like, I’m… the platform is something that we’ve iterated on for over a year, but it’s just a collection of tools. I even consider, like, our internal SOPs and things part of our Brainforge platform. It’s how we can take anybody off the street who comes in and becomes a Brainforge.
70 00:07:38.690 ⇒ 00:07:41.429 Uttam Kumaran: Engineer or employee and supercharge them.
71 00:07:41.510 ⇒ 00:07:47.650 Uttam Kumaran: You know, and it’s a mix of AI, it’s a mix of smart automation, it’s a mix of just organizing things better.
72 00:07:48.830 ⇒ 00:07:58.020 Uttam Kumaran: But for us, and the way I think about it as a business leader is, previously, I would have to go buy a Salesforce, or buy, like, a resource planning, or
73 00:07:58.020 ⇒ 00:08:10.160 Uttam Kumaran: large tool to go do this, and because of AI, we’ve actually been able to build our own software, pretty easily. And so, one is, like, I… we see the benefits. I use the platform every day. I want to continue to use it.
74 00:08:10.160 ⇒ 00:08:29.089 Uttam Kumaran: So, one is, like, okay, what’s… what’s… I want… it’s a little bit of a reflection on, like, okay, what’s been limiting us? The second thing is, like, okay, I think we’ve built a couple things, like, sort of halfway, but how do we get some of it out? And then third is, like, how do I promote so that more people in the company can get more ideas executed?
75 00:08:29.130 ⇒ 00:08:43.429 Uttam Kumaran: So part of this enablement plan is, like, I think there are a couple gatekeepers to ideas. I think I’m one of the big people, which is, like, I’m the one coming with a lot of ideas, so that needs to change. There’s a lot of smart people in the company, a lot of people with ideas about
76 00:08:43.429 ⇒ 00:09:03.160 Uttam Kumaran: automations that could save them time. The second thing is, okay, how does an idea actually get into code and get shipped? There’s a huge chain, a manufacturing chain there. Okay, so how do we break that down? How can we actually enable more people on the team to use AI to actually generate, like, a mock-up of what they want, or generate, like, a little animation of what they want?
77 00:09:03.160 ⇒ 00:09:17.069 Uttam Kumaran: Or even for people more technical, like you and me, Henry, can we even ship a first version just using cursor, and then hand it to the team to sort of, like, polish up? And that’s, like, what I… what I want to talk a little bit about today.
78 00:09:20.190 ⇒ 00:09:21.270 Henry Zhao: Okay, sounds good.
79 00:09:46.070 ⇒ 00:09:48.680 Casie Aviles: Yeah, I’m just reading through this talk.
80 00:09:49.420 ⇒ 00:09:49.810 Uttam Kumaran: No problem.
81 00:09:50.320 ⇒ 00:09:57.389 Casie Aviles: Yeah, curious, like, how we could, you know, potentially enable Non-technical makers, like…
82 00:09:58.860 ⇒ 00:10:06.020 Casie Aviles: Yeah, for, like, the UI stuff, I wonder how we could… potentially do that.
83 00:10:08.350 ⇒ 00:10:11.359 Uttam Kumaran: I’ve been playing around with over the last two weeks is, like.
84 00:10:11.540 ⇒ 00:10:17.400 Uttam Kumaran: I’ve been designing things in magic patterns, I’ve been trying
85 00:10:17.520 ⇒ 00:10:26.749 Uttam Kumaran: to just use AI to design things in Figma. So there is a lot of opportunity, actually, and I think we’re sitting on a lot of it, we just haven’t tried it yet.
86 00:10:27.760 ⇒ 00:10:37.030 Uttam Kumaran: You know, but I also think part of this is, like, there’s probably significant training that needs to happen, where we have to just train everyone in the company on how to use this stuff.
87 00:10:38.770 ⇒ 00:10:39.470 Casie Aviles: Yeah.
88 00:11:11.590 ⇒ 00:11:14.980 Uttam Kumaran: Okay, cool. Let me know when you guys are… Alrighty.
89 00:11:17.310 ⇒ 00:11:19.570 Uttam Kumaran: Please leave any comments.
90 00:12:29.290 ⇒ 00:12:30.250 Mustafa Raja: Oh, yeah, I’m good.
91 00:12:37.700 ⇒ 00:12:39.840 Uttam Kumaran: Okay, Casey, let me know when you’re ready.
92 00:12:41.300 ⇒ 00:12:46.559 Casie Aviles: Yeah, I’m just writing some comments, but yeah, I think we can dive in.
93 00:12:47.610 ⇒ 00:12:48.840 Uttam Kumaran: Okay, cool.
94 00:13:04.920 ⇒ 00:13:06.000 Uttam Kumaran: One sec…
95 00:13:17.710 ⇒ 00:13:20.119 Uttam Kumaran: Hmm, where is my screen?
96 00:13:28.830 ⇒ 00:13:30.680 Uttam Kumaran: Oh, it’s just the rainfall.
97 00:13:54.750 ⇒ 00:13:56.909 Uttam Kumaran: Okay, so…
98 00:13:57.170 ⇒ 00:14:14.089 Uttam Kumaran: there’s a couple of things, and I put this under just me and Sam, but this was gonna be my structure for our conversation anyway, so, one is, like, this is gonna be our sort of homepage for how AI gets used in the company. As you can see, I’m starting to build out
99 00:14:14.320 ⇒ 00:14:21.049 Uttam Kumaran: like, what are each of the core processes in the company? And then we’re gonna start to build out, okay, like.
100 00:14:21.280 ⇒ 00:14:28.029 Uttam Kumaran: how important it is, how important it would be to use AI to do it, who’s the customer. So we’re going to continue to build
101 00:14:28.210 ⇒ 00:14:33.390 Uttam Kumaran: this out. But also, down here, we have sort of a little bit of an audit of, like, what
102 00:14:33.520 ⇒ 00:14:34.660 Uttam Kumaran: exists.
103 00:14:34.940 ⇒ 00:14:38.640 Uttam Kumaran: Already. And so, this is something that I think we’re…
104 00:14:39.050 ⇒ 00:14:43.490 Uttam Kumaran: we’re… we’re gonna keep maintained. But really, today.
105 00:14:43.610 ⇒ 00:14:56.460 Uttam Kumaran: I kind of want to talk about this talk, which is, how do we expand the surface area of people that can contribute to the platform in both ideas and in actually shipping features?
106 00:14:57.980 ⇒ 00:15:17.960 Uttam Kumaran: And so, to talk about, like, my journey with this, like, I have no sort of formal background in full-stack engineering. I’ve managed product teams before, but my expertise is in data, but I’ve shipped a couple of features on the platform, in the last, like, 2 or 3 months, and I found it to be actually not as hard, and I actually think
107 00:15:18.190 ⇒ 00:15:34.659 Uttam Kumaran: We’re probably just a few iterations away from making it so anybody who’s even slightly technical can come on, think of an idea, and actually implement it. What do I mean when, I say, like, think of an idea? Let me…
108 00:15:34.670 ⇒ 00:15:40.479 Uttam Kumaran: Let me just share my… my whole screen, so, like, we can actually just have the platform up as I… as I talk.
109 00:15:49.280 ⇒ 00:15:50.250 Uttam Kumaran: Oh, okay.
110 00:15:57.550 ⇒ 00:16:01.879 Uttam Kumaran: Okay, let me know if this is, too big, but,
111 00:16:02.280 ⇒ 00:16:04.640 Uttam Kumaran: when I… when I talk about,
112 00:16:05.790 ⇒ 00:16:09.180 Uttam Kumaran: making a change on the platform, here’s one example.
113 00:16:09.310 ⇒ 00:16:13.379 Uttam Kumaran: So, in a given meeting.
114 00:16:13.590 ⇒ 00:16:19.439 Uttam Kumaran: I want to see these items, right? But I don’t like how they’re all tabs.
115 00:16:19.580 ⇒ 00:16:27.240 Uttam Kumaran: I think they should literally be separate, like, modules, where… because right now, if you look at this screen, there’s a lot of white space.
116 00:16:27.360 ⇒ 00:16:40.060 Uttam Kumaran: you have this giant thing here, you have this, but, like, all these things are helpful things that people may want to do. I think they should be split up. And so, okay, let’s… let’s build the user story, right? And so, what is a user story?
117 00:16:40.340 ⇒ 00:16:49.180 Uttam Kumaran: A user story is a concise description of a product feature’s need and value. As a type, as a blank, I want to blank so that I can achieve blank.
118 00:16:49.340 ⇒ 00:16:56.880 Uttam Kumaran: Right? So this is very famous in product management. Okay, so I would say as a, meeting attendee.
119 00:16:57.170 ⇒ 00:17:01.730 Uttam Kumaran: I want to… Clearly see what
120 00:17:02.120 ⇒ 00:17:04.300 Uttam Kumaran: sort of quick actions I can take.
121 00:17:04.319 ⇒ 00:17:07.899 Uttam Kumaran: given a meeting transcript, so that I can…
122 00:17:07.920 ⇒ 00:17:26.180 Uttam Kumaran: accomplish, like, my goal instead of manually doing more work, like creating an email summary, right? Another way is to say, hey, I want to, as a meeting, as a meeting attendee, I previously didn’t know you could do these things because they were hidden, and I want to know up front so that
123 00:17:26.640 ⇒ 00:17:37.560 Uttam Kumaran: I could start to use them, like, create linear tickets. Additionally, we are gonna probably have, like, 5 or 10 different things you can do out of a meeting. And so, I’m kind of going on a tangent right now, but basically.
124 00:17:37.710 ⇒ 00:17:50.099 Uttam Kumaran: I want to implement that feature. What is the typical state of the world? Okay, let me go make a ticket. Let me take a screenshot of this, make it, and say, I don’t like how this looks, I wish it was somewhere over here.
125 00:17:50.400 ⇒ 00:17:56.679 Uttam Kumaran: And then, okay, but let me tell you even what happens in a typical product org. I have to go to the product manager that owns this.
126 00:17:56.690 ⇒ 00:18:12.219 Uttam Kumaran: So let’s say that’s Rico. Rico says, great, thanks, okay, that’s the story. Let me go to design, and we’ll get a new design. Okay, let’s tweak for that. Okay, great, once it’s designed, let’s hand it to the AI team to build. Okay, they work on it among all their other priorities. Finally, it gets shipped.
127 00:18:12.410 ⇒ 00:18:23.400 Uttam Kumaran: it’s probably, like, not exactly what I wanted, but I’m just happy as a… as a user that it got out at all. And so, how do we cut that by… into, like, 10%?
128 00:18:23.550 ⇒ 00:18:29.760 Uttam Kumaran: So, what if I told you that I can actually… have this problem.
129 00:18:30.190 ⇒ 00:18:31.889 Uttam Kumaran: I can articulate it.
130 00:18:32.100 ⇒ 00:18:35.089 Uttam Kumaran: I can generate what a design of it could look like.
131 00:18:36.040 ⇒ 00:18:40.379 Uttam Kumaran: I can then take that design… I can continue to tweak that design until it’s something that I like.
132 00:18:40.790 ⇒ 00:18:46.369 Uttam Kumaran: I can then pass that design to the AI team to say, hey, is this something that we’d be happy with?
133 00:18:46.410 ⇒ 00:19:00.870 Uttam Kumaran: I can then actually take that design, pass it again to AI, and pass it to AI and have it actually do the first version of it. Okay, I may not get the… I may not get a couple of linings right, I may have messed up some APIs.
134 00:19:00.940 ⇒ 00:19:14.880 Uttam Kumaran: You know, I may not have considered broader architecture or file naming conventions, but I still probably got us 60% of the way there. At that point, I could then hand it over to the AI team to do the last mile of the last mile.
135 00:19:15.160 ⇒ 00:19:23.590 Uttam Kumaran: And so that’s the world that I want to live in as a semi-technical person here at Brainforge. I want to be able to look at the platform, say, I want this.
136 00:19:23.710 ⇒ 00:19:27.039 Uttam Kumaran: Propose the design, Get the proof on design.
137 00:19:27.140 ⇒ 00:19:29.759 Uttam Kumaran: Propose the change, and then…
138 00:19:30.030 ⇒ 00:19:39.479 Uttam Kumaran: basically propose the change as, like, a PR, or in a branch, and then hand it off to the team. So my question to this crew is, why can’t we do that today?
139 00:19:39.670 ⇒ 00:19:44.059 Uttam Kumaran: Right? That is the question I’m proposing, right? Why is that not possible today?
140 00:19:44.180 ⇒ 00:19:50.100 Uttam Kumaran: Does anyone have, like, any… Thoughts on that, or, like, any clarifying questions on that?
141 00:19:50.530 ⇒ 00:19:51.979 Uttam Kumaran: That state of the world.
142 00:20:00.570 ⇒ 00:20:05.950 Casie Aviles: You know, I think maybe one… one reason why is, you know, peop… I guess…
143 00:20:06.560 ⇒ 00:20:09.589 Casie Aviles: Like, in terms of, like, a delegation of jobs, like…
144 00:20:10.240 ⇒ 00:20:16.220 Casie Aviles: of tasks, like, for the AI team, or just, you know, kind of expected to just do it.
145 00:20:17.210 ⇒ 00:20:23.369 Casie Aviles: from… From a ticket, and then we will just, work on that.
146 00:20:23.570 ⇒ 00:20:25.170 Casie Aviles: From scratch, I guess.
147 00:20:26.180 ⇒ 00:20:32.870 Casie Aviles: like… I guess the people who have, like, an idea don’t have…
148 00:20:34.310 ⇒ 00:20:42.179 Casie Aviles: a sense of how they could… they could also start, working on it, you know? Like, I guess training, that’s one thing, maybe.
149 00:20:43.580 ⇒ 00:20:45.820 Uttam Kumaran: Yeah, and also, I think,
150 00:20:46.080 ⇒ 00:20:57.680 Uttam Kumaran: Let me… maybe if I was to ask you, Mustafa, tell me, like, what kind of constraints we’re working with on the AI team? Like, if you… if you were to name a couple. Like, what are our… what are our constraints?
151 00:21:00.040 ⇒ 00:21:04.970 Mustafa Raja: Our constraints in terms of enabling.
152 00:21:05.480 ⇒ 00:21:16.349 Uttam Kumaran: No, no, just in general, like, let’s take the example that I said, where we want to ship that feature, right? What are the constraints that we’re… we’re dealing with on the AI team that may impact
153 00:21:16.550 ⇒ 00:21:24.520 Uttam Kumaran: our ability to deliver on that within a timeframe, and accurately, and the best. Like, think about what are our constraints that we’re dealing with.
154 00:21:24.670 ⇒ 00:21:31.360 Mustafa Raja: Yeah, so usually, what we do is we do not go, to the design team.
155 00:21:31.670 ⇒ 00:21:41.069 Mustafa Raja: That would be for the features that I have shipped. Most of them wouldn’t have a design pre-decided.
156 00:21:41.760 ⇒ 00:21:48.000 Mustafa Raja: So I just go ahead and, work with Cursor and, and implement those, so…
157 00:21:48.000 ⇒ 00:22:03.070 Mustafa Raja: In terms of this, I don’t think that there are much, constraints. The semi-technical people, to get this out to them, I think a workshop where we actually,
158 00:22:03.070 ⇒ 00:22:09.619 Uttam Kumaran: No, no, but I guess you’re jumping too far, because I just want to say the obvious out loud. So, one constraint you hit on is…
159 00:22:10.070 ⇒ 00:22:13.789 Uttam Kumaran: We don’t have the time or the resources to get everything designed.
160 00:22:14.220 ⇒ 00:22:14.940 Mustafa Raja: Yeah.
161 00:22:14.940 ⇒ 00:22:15.640 Uttam Kumaran: Right.
162 00:22:15.700 ⇒ 00:22:17.839 Mustafa Raja: Yeah. The second piece is…
163 00:22:18.720 ⇒ 00:22:24.659 Uttam Kumaran: Let’s say we had… Let’s say we had a backlog of things that need to get done.
164 00:22:24.910 ⇒ 00:22:28.770 Uttam Kumaran: like we do. What is our constraint right now?
165 00:22:29.220 ⇒ 00:22:32.009 Uttam Kumaran: Sorry, it’s very obvious. What is it?
166 00:22:33.370 ⇒ 00:22:35.050 Henry Zhao: Human hours.
167 00:22:35.430 ⇒ 00:22:36.320 Uttam Kumaran: Time.
168 00:22:36.840 ⇒ 00:22:38.340 Uttam Kumaran: It’s pure time.
169 00:22:39.010 ⇒ 00:22:39.900 Henry Zhao: Exactly.
170 00:22:40.370 ⇒ 00:22:43.949 Uttam Kumaran: And so, as Henry said, our constraint is time.
171 00:22:44.110 ⇒ 00:22:50.810 Uttam Kumaran: The third constraint is what I would… what I would call fidelity, or really true understanding of the problem.
172 00:22:51.020 ⇒ 00:23:01.349 Uttam Kumaran: as the… let’s say I’m putting my user hat on, right? It’s really clear to me what the problem is. I go to a meeting, and I have to click 4 buttons to get a follow-up email.
173 00:23:01.480 ⇒ 00:23:12.990 Uttam Kumaran: There should just be a thing right here, or I should just scroll down and get it, because maybe I want to get the follow you, and I want to do summary, right? I understand what the pain is, but what happens if you guys have ever played this game telephone?
174 00:23:13.210 ⇒ 00:23:14.170 Uttam Kumaran: Right? Telephone.
175 00:23:14.170 ⇒ 00:23:14.509 Casie Aviles: I was like.
176 00:23:14.510 ⇒ 00:23:23.419 Uttam Kumaran: you’re sitting in a circle, you tell someone something, then they go around, and by the time it gets to the end, it’s completely different, right? This is the exact same problem that happens in engineering.
177 00:23:23.550 ⇒ 00:23:39.340 Uttam Kumaran: And what you’re gonna see is I’m going to describe a problem. This happens on all of our engineering teams, but I’m just focusing on the AI team, is once I translate… once I take this and I take my ideas, and I give it to Rico, Ricoh is gonna go summarize it into a ticket.
178 00:23:39.480 ⇒ 00:23:43.350 Uttam Kumaran: Immediately, there’s a loss factor, right?
179 00:23:43.480 ⇒ 00:23:49.699 Uttam Kumaran: There is no way that that ticket, given our situation, has everything that I mentioned.
180 00:23:49.870 ⇒ 00:23:55.750 Uttam Kumaran: And then what happens? When that email… when that ticket goes to the designer, There’s another loss.
181 00:23:55.930 ⇒ 00:24:00.209 Uttam Kumaran: And then when the designer comes back and gives it to us to review.
182 00:24:00.700 ⇒ 00:24:05.290 Uttam Kumaran: Now they’re… now we’re stuck with a design, because guess what? We can only do one or two iterations.
183 00:24:05.440 ⇒ 00:24:17.030 Uttam Kumaran: So this is why, when you talk about why is… why is software designed the way it is, it’s because design is commonly not given enough time to design, right? Because they would take
184 00:24:17.120 ⇒ 00:24:19.870 Uttam Kumaran: They need, like, multiple iterations to get perfect.
185 00:24:19.910 ⇒ 00:24:31.080 Uttam Kumaran: that’s why most software is really shitty. It’s because they… there’s just no way to spend enough time doing the design work necessary. But then you have great software. You have companies like Shopify, you have…
186 00:24:31.080 ⇒ 00:24:47.099 Uttam Kumaran: tools that I use, like VimCal, right, for example, or YouTube, where they have 100,000 designers. They’re gonna design great products, but we don’t have that luxury. And so we’re in a design-constrained environment. We’re also in a time-constrained environment.
187 00:24:47.280 ⇒ 00:24:53.909 Uttam Kumaran: But what… what… what don’t… what aren’t we constrained by? Well, we have a really smart people, right? So we’re not constrained by aptitude.
188 00:24:54.140 ⇒ 00:25:10.150 Uttam Kumaran: Second is we’re not constrained by, what has changed, though, is all those things could have been true 5 years ago, but right now, we have a solution for it. We actually… there are tools in the market that can help me, as the user, design the iteration of it.
189 00:25:10.200 ⇒ 00:25:24.700 Uttam Kumaran: and actually ship, like, a first PR, which cuts in a couple ways. One, it solves our time constraint. The AI team and their limited time, which is… I’m telling you, we’re never gonna solve that, guys. It’s all… we’re always gonna have limited time.
190 00:25:24.820 ⇒ 00:25:35.689 Uttam Kumaran: So it solves that problem. Second is it solves a design constraint problem, where design is the only… they’re the only people that can, like, mock up something like this. To give you, like.
191 00:25:36.030 ⇒ 00:25:43.830 Uttam Kumaran: again, like, we’ll head back to the doc in a sec, but to give you, like, sort of… this is just one of these tools, but I’ve been using this tool, right? Like.
192 00:25:44.000 ⇒ 00:25:50.649 Uttam Kumaran: I… I wanted to check out, like, what a new version of the platform could look like, and
193 00:25:51.420 ⇒ 00:25:54.899 Uttam Kumaran: Let me show you a couple examples of, like, what I generated.
194 00:25:55.030 ⇒ 00:25:59.960 Uttam Kumaran: So, my first version of this, was…
195 00:26:00.400 ⇒ 00:26:03.840 Uttam Kumaran: I want to reimagine the homepage of the platform.
196 00:26:04.450 ⇒ 00:26:11.249 Uttam Kumaran: like… We have meetings with our clients and a general chat, and then there’s a specific meeting view.
197 00:26:11.380 ⇒ 00:26:14.460 Uttam Kumaran: Right? Okay, this is what the new design is.
198 00:26:14.820 ⇒ 00:26:28.829 Uttam Kumaran: all right, looks kind of, like, pretty good, right? And you can click, and then there’s nothing there, right? So, it went ahead and… the nice thing about this tool is it creates some of this stuff, but it doesn’t do everything. And so, let me go,
199 00:26:29.150 ⇒ 00:26:31.049 Uttam Kumaran: Let me go back…
200 00:26:33.290 ⇒ 00:26:39.609 Uttam Kumaran: And so then I said, great, I want you to do a design critique and suggest some changes.
201 00:26:39.790 ⇒ 00:26:44.099 Uttam Kumaran: I then gave it our AI goals, our OKRs.
202 00:26:44.330 ⇒ 00:26:49.529 Uttam Kumaran: I gave her that entire doc, and I said, here’s everything about our AI platform.
203 00:26:49.640 ⇒ 00:26:51.970 Uttam Kumaran: I said, okay, I thought about it, I could see that
204 00:26:52.080 ⇒ 00:26:54.889 Uttam Kumaran: We’re trying to create this powerful, ambitious platform.
205 00:26:55.000 ⇒ 00:26:57.129 Uttam Kumaran: It then gave me another… another view.
206 00:26:57.370 ⇒ 00:26:59.129 Uttam Kumaran: And so, this is the V2.
207 00:27:00.310 ⇒ 00:27:04.660 Uttam Kumaran: I didn’t do anything. I literally copy-pasted that notion.
208 00:27:05.550 ⇒ 00:27:09.160 Uttam Kumaran: this… this AI platform notion, pasted in here and said.
209 00:27:09.360 ⇒ 00:27:12.369 Uttam Kumaran: This is what we’re trying to accomplish, and look how it looks.
210 00:27:15.650 ⇒ 00:27:18.950 Uttam Kumaran: It’s auto… and let me point out a couple of interesting features.
211 00:27:19.460 ⇒ 00:27:23.189 Uttam Kumaran: there’s icons, For the different co- for the different things.
212 00:27:23.430 ⇒ 00:27:28.320 Uttam Kumaran: It’s able to tag… there’s tags, there’s some sort of tagging mechanism.
213 00:27:28.600 ⇒ 00:27:33.999 Uttam Kumaran: There’s ability to show and hide filters, there’s some business KPIs at the top.
214 00:27:34.160 ⇒ 00:27:38.960 Uttam Kumaran: I don’t think it went ahead and did anything on the client side.
215 00:27:39.110 ⇒ 00:27:41.560 Uttam Kumaran: But…
216 00:27:42.160 ⇒ 00:27:52.119 Uttam Kumaran: it’s decent, right? It’s like a decent start at something, and I didn’t even tell it anything. So then my immediate question is, okay, left bar is pretty bare, like, can you reimagine the left bar?
217 00:27:52.260 ⇒ 00:27:55.140 Uttam Kumaran: Great. Here’s, like, what it came up with.
218 00:27:57.940 ⇒ 00:28:04.879 Uttam Kumaran: all… but I… I just… what I want to show you guys is that’s… I asked… I literally typed in
219 00:28:05.250 ⇒ 00:28:10.100 Uttam Kumaran: This trash prompt… And we got this beautiful thing out.
220 00:28:10.980 ⇒ 00:28:18.130 Uttam Kumaran: And what I’m trying to share is I’ve… I’ve immediately attacked two fundamental constraints. Time.
221 00:28:18.610 ⇒ 00:28:23.060 Uttam Kumaran: Right? So now I’ve gone, as someone who… I have no design background.
222 00:28:23.390 ⇒ 00:28:34.339 Uttam Kumaran: I’m able… and I… and I even gave it… I didn’t even give it any guidance. I gave it 3 sentences. Imagine if, Casey, it was, like, one… something so small as, like, I want to redesign how the filters look.
223 00:28:35.850 ⇒ 00:28:43.959 Uttam Kumaran: And I gave it all of my problems on why I don’t like the filters, and I said, okay, give me a new version. It’s gonna nail it.
224 00:28:44.190 ⇒ 00:28:49.349 Uttam Kumaran: And then what I can do, I can immediately take that, and guess what, lovely this gives us?
225 00:28:49.680 ⇒ 00:28:53.410 Uttam Kumaran: TypeScript files, I can then hand it to you guys to go push through.
226 00:28:53.800 ⇒ 00:29:01.529 Uttam Kumaran: Better yet, given I’m slightly technical and I use Cursor, I can actually just take a screenshot of this and throw it right into Cursor.
227 00:29:02.490 ⇒ 00:29:10.120 Uttam Kumaran: And so what I want to show you is, like, here’s a demo of where I’ve immediately broken those two assumptions. Everybody can now design.
228 00:29:10.450 ⇒ 00:29:11.230 Uttam Kumaran: Right?
229 00:29:11.390 ⇒ 00:29:22.569 Uttam Kumaran: And everybody now has the ability to also go and develop at least a proof of concept of something in code. And so this is sort of what I want to enable everybody in the company to do.
230 00:29:23.800 ⇒ 00:29:26.769 Uttam Kumaran: beginning, I want to really enable, with our platform.
231 00:29:27.020 ⇒ 00:29:38.490 Uttam Kumaran: But then, of course, I want this to be something that we can do for clients as well. Using these design tools and using our coding tools to speed up the process for them.
232 00:29:38.630 ⇒ 00:29:43.749 Uttam Kumaran: But, like, That’s basically what I wrote in this document. It’s like…
233 00:29:44.340 ⇒ 00:29:48.750 Uttam Kumaran: is that? I have… I made some tool-specific bets, but, like.
234 00:29:49.300 ⇒ 00:29:52.809 Uttam Kumaran: that’s, like, the meat of, like, what I… what I wrote about, so I’ll pause there.
235 00:29:53.190 ⇒ 00:29:58.440 Henry Zhao: Yeah, this is great. My main concern, I would say, is, like, adoption and trust.
236 00:29:58.650 ⇒ 00:30:07.230 Henry Zhao: Like, I would be very worried that, am I using this properly? So I wonder if we address that with, I don’t know, whether it’s, like, training or…
237 00:30:07.740 ⇒ 00:30:09.419 Henry Zhao: Something like that, I don’t know.
238 00:30:09.420 ⇒ 00:30:15.270 Uttam Kumaran: Can you go deeper into that? Like, what… what… where would you feel nervous about what parts in particular?
239 00:30:15.620 ⇒ 00:30:29.280 Henry Zhao: Yeah, I’m just thinking, like, as the end user who’s maybe not so well-versed in AI, I don’t know if me saying, like, any imagination of the left bar is enough to give it what I need to meet my business needs, you know what I mean?
240 00:30:29.660 ⇒ 00:30:32.010 Henry Zhao: So then I might be afraid to adopt it.
241 00:30:32.160 ⇒ 00:30:44.700 Henry Zhao: So it might be, like, I will do the demo, I’ll be like, this looks great, but then I won’t actually adopt it, because deep down, I have this fear of, like, it’s not actually doing what I need it to do, and I’ll… I’ll deal with that maybe later on, when I have more time, and then I never have that time.
242 00:30:45.310 ⇒ 00:30:46.380 Uttam Kumaran: I see, okay.
243 00:30:47.040 ⇒ 00:30:49.539 Henry Zhao: I’m just bringing that up as, like, a potential, I think, roadblock.
244 00:30:49.540 ⇒ 00:30:50.489 Uttam Kumaran: No, it’s fair.
245 00:30:50.860 ⇒ 00:30:51.630 Uttam Kumaran: Yeah.
246 00:30:52.770 ⇒ 00:31:02.780 Henry Zhao: Because I think this is one of those, like, too-good-to-be-true things, where it’s like, if it works really well, then I think some of the later adopters go into it with a little bit more hesitation. Almost like a double-edged sword.
247 00:31:03.760 ⇒ 00:31:09.150 Uttam Kumaran: Yeah, so the one thing I’ll talk about is, one, everybody in the company you know…
248 00:31:09.210 ⇒ 00:31:11.820 Uttam Kumaran: This is one where we are the client.
249 00:31:11.870 ⇒ 00:31:25.429 Uttam Kumaran: where everybody in the company, I think, is pretty pro-AI. I haven’t heard any immediate pushback, but I also do know that everybody is upskilled a lot. And so this is another area where we have to upskill. But I also think it’s helpful for everybody to get a win.
250 00:31:25.430 ⇒ 00:31:34.020 Uttam Kumaran: Imagine we had a meeting where I said, everybody go and think about a feature that you want to add to the platform, and by the end of this meeting, it will make it into the product.
251 00:31:35.820 ⇒ 00:31:39.030 Uttam Kumaran: you’re gonna say, no way. And I’m gonna say, okay, let’s do it.
252 00:31:39.720 ⇒ 00:31:47.259 Uttam Kumaran: Everybody start with the user story, but this is where, like, what you can’t speed up is the product
253 00:31:47.410 ⇒ 00:31:52.080 Uttam Kumaran: development fundamentals. You need to have a clearly articulated problem.
254 00:31:52.300 ⇒ 00:31:58.600 Uttam Kumaran: Like, you can’t just say, make this better. Like, what I did is not good. I just needed to demo, I just needed to test something.
255 00:31:58.810 ⇒ 00:32:01.869 Uttam Kumaran: And figure out, like, okay, instead, like.
256 00:32:02.050 ⇒ 00:32:07.049 Uttam Kumaran: You need to have a clear problem you’re trying to solve. You need to treat it like a product, like a…
257 00:32:07.180 ⇒ 00:32:11.150 Uttam Kumaran: product designer and an engineer, you need to say, what is the problem I’m trying to solve?
258 00:32:11.440 ⇒ 00:32:20.620 Uttam Kumaran: okay, here’s, like, what I… here’s an idea of what I think it could do, and then you get the designs. You then have to discuss with people, hey, is this… is this worth it? Is this, like…
259 00:32:20.740 ⇒ 00:32:29.629 Uttam Kumaran: Is this something you guys are already working on somewhere else? Okay, great, we’re good with the design. Then you run… then you run it through Cursor, and then again, there’s a human in the loop.
260 00:32:29.630 ⇒ 00:32:44.070 Uttam Kumaran: the AI team has to take it to get it deployed. So, it’s actually… you still have to do the pieces that matter, which I think fundamentally is the fact of thinking through the problem, and thinking through scenarios of why this is a problem.
261 00:32:44.270 ⇒ 00:32:52.809 Uttam Kumaran: Everything right now in AI is limited by shitty context. If you give it very bad context, you’re gonna get a scattered output.
262 00:32:53.020 ⇒ 00:32:57.090 Uttam Kumaran: The great thing about a tool like this is they took my crappy
263 00:32:57.210 ⇒ 00:33:00.390 Uttam Kumaran: Input, and they still did a decent job.
264 00:33:00.570 ⇒ 00:33:03.890 Uttam Kumaran: Right? And… but the problem is, like, this is where…
265 00:33:04.210 ⇒ 00:33:06.500 Uttam Kumaran: You’re right in that we have to train
266 00:33:06.630 ⇒ 00:33:11.399 Uttam Kumaran: To say, give it a lot of context, and spend more time on the context.
267 00:33:11.540 ⇒ 00:33:13.090 Henry Zhao: Yeah. Than you think.
268 00:33:13.090 ⇒ 00:33:13.650 Uttam Kumaran: I guess.
269 00:33:13.650 ⇒ 00:33:18.450 Henry Zhao: I just need… I just need time to, like, do a few examples so that I know that the context I’m.
270 00:33:18.450 ⇒ 00:33:18.980 Uttam Kumaran: It is amazing.
271 00:33:18.980 ⇒ 00:33:19.730 Henry Zhao: stuff.
272 00:33:19.840 ⇒ 00:33:21.180 Henry Zhao: And I probably need to.
273 00:33:21.180 ⇒ 00:33:24.800 Uttam Kumaran: Part of it is learning how to Google again. It’s similar to that, you know?
274 00:33:26.050 ⇒ 00:33:26.740 Henry Zhao: Right.
275 00:33:27.450 ⇒ 00:33:33.840 Henry Zhao: Like, this tool, I would probably need some time to kind of, and I wouldn’t say test it out, but yeah, just get a better understanding.
276 00:33:35.040 ⇒ 00:33:35.949 Uttam Kumaran: Makes sense.
277 00:33:35.950 ⇒ 00:33:36.490 Henry Zhao: Yeah.
278 00:33:43.150 ⇒ 00:33:47.899 Henry Zhao: Just as another example, right? Like, the voice input stuff, like the BrainForge meeting tool.
279 00:33:47.900 ⇒ 00:33:48.340 Uttam Kumaran: Yes.
280 00:33:48.340 ⇒ 00:33:52.989 Henry Zhao: is Abigail Zhao, so that’s, like, one piece of, like, trust erosion that just, like, over time.
281 00:33:52.990 ⇒ 00:33:53.770 Casie Aviles: It’s like…
282 00:33:53.940 ⇒ 00:33:55.329 Henry Zhao: I need to evaluate, you know?
283 00:33:56.640 ⇒ 00:34:06.180 Uttam Kumaran: So that’s an exact… yeah, but see, this is the thing, is like, I want to make it seem like you can take that piece of feedback and actually get it fixed.
284 00:34:06.360 ⇒ 00:34:21.740 Uttam Kumaran: So part of this is just, like… see, I know Casey heard that, he’s like, shit, what the hell? Like, I know exactly, like, because we probably hard-coded something where Abigail’s in there, in Superbase somewhere, right? But, like, we can solve that, but it has to be, like, you as a user have to feel confident that
285 00:34:22.070 ⇒ 00:34:33.409 Uttam Kumaran: you could submit a change, and then you’ll get it fixed. And for that, that’s like a, hey, I saw a mistake here, but let’s say you’re using the platform, and you’re like, hey, I was using, like.
286 00:34:33.489 ⇒ 00:34:45.510 Uttam Kumaran: the linear tickets agent, and I just don’t like this UI, I want to propose a different one. You can now come to the table not only with the problem, but even, like, a draft version of, like, what you think could work.
287 00:34:45.730 ⇒ 00:34:50.249 Uttam Kumaran: Which puts you as a user in a much more… in a much more equipped spot.
288 00:34:50.350 ⇒ 00:34:51.270 Henry Zhao: Right?
289 00:34:51.270 ⇒ 00:35:00.920 Uttam Kumaran: Versus, like, I have this problem, like, I don’t know how to solve it, you know? I actually want to give more power to the end users to dictate what the platform is for them, because
290 00:35:01.400 ⇒ 00:35:03.269 Uttam Kumaran: We’re the customers, you know?
291 00:35:03.690 ⇒ 00:35:04.329 Henry Zhao: Yeah, and then.
292 00:35:04.330 ⇒ 00:35:07.039 Uttam Kumaran: Like, Henry, you are the customer for this platform.
293 00:35:08.000 ⇒ 00:35:27.220 Henry Zhao: Yeah, I would say another concern I had earlier when we were talking about constraints about, like, human time and everything, another one, I think, is just so many different use cases, so as we grow and as we scale, I think we’re gonna start having people with different use cases and different ways they interact with AI and different ways they use the content, where what works for one group of people might not be, like.
294 00:35:27.410 ⇒ 00:35:35.690 Henry Zhao: obvious or work for a different group of people, that has, like, different search habits or different usage habits, and it’s something that I think we should also just keep in mind as we grow and scale.
295 00:35:36.950 ⇒ 00:35:48.980 Uttam Kumaran: No, you’re totally right, and this is where I actually… I want this stuff, I think there’ll be some stuff that’s in the platform, but I also think some stuff needs to happen in Slack. I also still… I use the… just…
296 00:35:49.160 ⇒ 00:35:58.629 Uttam Kumaran: I just use ChatGPT stuff every day, like the desktop app. So I don’t… this is where, like, I don’t care as much as, like.
297 00:35:58.880 ⇒ 00:36:02.550 Uttam Kumaran: people use the right thing. I care that people are using it.
298 00:36:02.690 ⇒ 00:36:13.790 Uttam Kumaran: And that we are taking advantage of this, like, incredible technology that got dropped in our lap, and we build, like, an incredible business that way. And so, part of this is, like.
299 00:36:14.340 ⇒ 00:36:22.710 Uttam Kumaran: the AI team is the… should be, like, the governor over all of the architecture, but I don’t want them to be the gatekeeper of ideas.
300 00:36:22.970 ⇒ 00:36:27.370 Uttam Kumaran: And I don’t want them to be the gatekeeper of ideas making it to production.
301 00:36:27.540 ⇒ 00:36:38.500 Uttam Kumaran: You know, I want… I want other people to contribute to ideas making it all the way. You know, because usually this is, again, this is what happens, it’s like, I don’t think… I don’t think…
302 00:36:38.650 ⇒ 00:36:43.840 Uttam Kumaran: If… for the amount of backlog we have, I should go hire 10 people, but we can’t do that.
303 00:36:43.950 ⇒ 00:36:48.500 Uttam Kumaran: Right? So there has to be an alternative, and I think we… we have a few that we can work on.
304 00:36:48.890 ⇒ 00:36:51.260 Henry Zhao: Yeah, that part is clear, and that part I agree with.
305 00:36:56.060 ⇒ 00:37:02.710 Uttam Kumaran: So that’s… that’s sort of, like, how I’m thinking. I mean, another piece of this that I wrote is just, like, again, thinking about
306 00:37:02.960 ⇒ 00:37:17.040 Uttam Kumaran: how do we improve prompts, and the amount, the richness of prompts, is voice. One thing that I use daily, I use Whisper, you see here on the bottom of my screen, and if I pull up Whisper.
307 00:37:17.230 ⇒ 00:37:19.459 Uttam Kumaran: I’ve been using it all day for stuff.
308 00:37:19.600 ⇒ 00:37:24.060 Uttam Kumaran: And… basically…
309 00:37:24.200 ⇒ 00:37:36.309 Uttam Kumaran: I use it any… when I… when I go into ChatGPT, I may have to type in something that’s, like, super, super large, and I don’t want to type all this out. Like, look how much I have to type.
310 00:37:36.430 ⇒ 00:37:41.369 Uttam Kumaran: But this is what I need to tell AI to get a really rich output. And so I just speak it.
311 00:37:41.790 ⇒ 00:37:53.290 Uttam Kumaran: And so, you can see I’ve, like, spoken a ton of words, and anytime I’m, like, in here and I need to leave a large comment, or I need to give AI a lot of rich context.
312 00:37:53.470 ⇒ 00:37:58.110 Uttam Kumaran: instead of typing it, which is gonna cause me to just give it limited, I just talk.
313 00:37:58.240 ⇒ 00:38:07.309 Uttam Kumaran: And I can just talk for 30 seconds, give it everything it needs. And so that’s something that… it’s working for me, and I want to promote so that everybody can do that.
314 00:38:07.530 ⇒ 00:38:17.760 Uttam Kumaran: But everybody on their own journey has to try it and find it out, but again, right now, it’s not available. You can’t type into here. You can’t type into our… our chat.
315 00:38:18.150 ⇒ 00:38:24.950 Uttam Kumaran: our chat systems at all. You can’t… you can’t speak into them at all, right? So that’s something that I want to enable, but here… but let me give you an example.
316 00:38:24.990 ⇒ 00:38:44.229 Uttam Kumaran: I have to then… even for this, this is a bit of a mega meta example. I have to go to Ricoh, create a ticket, I have to talk to Sam and the ad team, we have to decide on the best speech-to-text SDK. We then have to… someone on the team has to take that, develop it, run it by me for feedback, run it by PR review.
317 00:38:44.780 ⇒ 00:38:57.040 Uttam Kumaran: I’m gonna do, like, later this week, I already have… either we’re gonna use 11 Labs, or I think we can use OpenAI’s real-time voice API, or something. I’m gonna just do the first version.
318 00:38:57.110 ⇒ 00:39:07.039 Uttam Kumaran: I’m gonna get it working locally, and then I’m gonna say, hey guys, I got this working locally, I don’t know if it’s perfect, but can I hand this branch off and someone can finish the tackle?
319 00:39:07.920 ⇒ 00:39:09.989 Uttam Kumaran: I’ve just cut, like, 10 steps.
320 00:39:10.140 ⇒ 00:39:19.209 Uttam Kumaran: You know? That’s what I want to… that’s what I want to enable. But I can’t be the only person in the company that can do that. Like, I want anyone to be able to…
321 00:39:19.390 ⇒ 00:39:21.770 Uttam Kumaran: Get to that point where they can hand…
322 00:39:22.060 ⇒ 00:39:25.809 Uttam Kumaran: like, almost a partially finished branch to the AI team to wrap up.
323 00:39:26.140 ⇒ 00:39:28.549 Uttam Kumaran: that’s the real, like, holy grail, I think.
324 00:39:31.320 ⇒ 00:39:32.930 Henry Zhao: Okay, sounds good, looks good.
325 00:39:35.430 ⇒ 00:39:41.910 Casie Aviles: I think, you know, people being able to, like, at least contribute in that way, that could also…
326 00:39:42.550 ⇒ 00:39:51.749 Casie Aviles: Help with adoption, because… They… I guess they have more ownership or investment to the outcome, but yeah.
327 00:39:53.300 ⇒ 00:39:54.350 Uttam Kumaran: 100%.
328 00:39:58.890 ⇒ 00:40:00.349 Uttam Kumaran: Mustafa, what do you think?
329 00:40:01.480 ⇒ 00:40:03.490 Mustafa Raja: Yeah, I… I agree.
330 00:40:03.680 ⇒ 00:40:10.429 Mustafa Raja: Getting people to work with this would definitely, increase.
331 00:40:11.750 ⇒ 00:40:13.959 Mustafa Raja: Usage of platform.
332 00:40:16.180 ⇒ 00:40:21.249 Uttam Kumaran: So I think two things that our team, AI team, needs to figure out is, one.
333 00:40:22.010 ⇒ 00:40:29.349 Uttam Kumaran: like, whether we’re gonna use a tool like Magic Patterns, or a tool like Lovable, to enable the…
334 00:40:30.080 ⇒ 00:40:31.660 Uttam Kumaran: Text to design.
335 00:40:32.520 ⇒ 00:40:37.240 Uttam Kumaran: Magic Patterns has been pretty good. There is a direct Figma
336 00:40:37.480 ⇒ 00:40:41.969 Uttam Kumaran: integration. So, like, I uploaded all of our components and stuff.
337 00:40:42.210 ⇒ 00:40:53.939 Uttam Kumaran: So, one is, like, I think we need to decide as a team what is gonna be our design, like, our AI design tool of choice, and we need to enable that tool. So, whether it’s
338 00:40:54.270 ⇒ 00:40:57.050 Uttam Kumaran: plugged into our Figma files, or whatever.
339 00:40:57.300 ⇒ 00:41:03.989 Uttam Kumaran: That’s one thing. The second thing is cursor. So even yesterday, I was, like, trying to ship something, but I didn’t have the end keys on my laptop.
340 00:41:04.300 ⇒ 00:41:13.980 Uttam Kumaran: Okay, so we have to think about, like, what is a user experience of shipping a feature here at Brainforge? And so we have to make sure that everyone’s cursor is set up.
341 00:41:14.110 ⇒ 00:41:16.289 Uttam Kumaran: It’s really clear how they can…
342 00:41:16.460 ⇒ 00:41:23.800 Uttam Kumaran: Use the cursor plan mode to plan out a feature, to create a branch, They need to know, like.
343 00:41:23.980 ⇒ 00:41:40.679 Uttam Kumaran: And then, can we have some type of PR review process? So, those are the things that I think our team needs to work on, because I have a million ideas, and I think Rico does, Ryan does, but they’re limited by… like, we’re limited by the fact that I can’t write TypeScript. I’ve never written a line of TypeScript in my life.
344 00:41:41.150 ⇒ 00:41:49.040 Uttam Kumaran: Yet, like, I can debug TypeScript probably half-decently. I can certainly use Cursor.
345 00:41:49.530 ⇒ 00:41:56.030 Uttam Kumaran: So, how do you enable a platform like this for me to develop on? You know, that’s what I think we need to think about.
346 00:41:56.300 ⇒ 00:42:01.589 Uttam Kumaran: I think if you look at the history of our team, I think we’ve always had these constraints, and that’s limited us, because
347 00:42:01.830 ⇒ 00:42:09.690 Uttam Kumaran: you know, Casey, like, we’ve only had a couple people on the team, and there’s so much we want to build, and now that you guys are working on clients, there’s even less time
348 00:42:09.810 ⇒ 00:42:21.289 Uttam Kumaran: And so, that’s something that I think we should try to focus more on, is building, like, a true ecosystem where anyone in the company can be a designer or an engineer on the platform.
349 00:42:21.760 ⇒ 00:42:25.840 Uttam Kumaran: And I think that’ll accelerate the amount of features that we get out, for sure.
350 00:42:26.050 ⇒ 00:42:29.220 Uttam Kumaran: And we’ll actually get to cruise through our backlog a little bit faster.
351 00:42:37.330 ⇒ 00:42:38.170 Uttam Kumaran: Cool.
352 00:42:38.350 ⇒ 00:42:42.139 Uttam Kumaran: So, I think I’m gonna sort of talk to,
353 00:42:42.280 ⇒ 00:42:46.930 Uttam Kumaran: Sam about this a little bit. I still think, like, continue course this week.
354 00:42:47.110 ⇒ 00:42:57.760 Uttam Kumaran: But, I would like to see us make some movements on choosing our design, Tool of choice… And…
355 00:42:58.790 ⇒ 00:43:03.549 Uttam Kumaran: having a very clear SOP on how to develop and push a PR,
356 00:43:03.700 ⇒ 00:43:09.249 Uttam Kumaran: and cursor on our repo. I know we already… I implemented cursor rules, we have a couple things, but…
357 00:43:09.490 ⇒ 00:43:17.919 Uttam Kumaran: I think that’s a really good use for time, and then that way, you, Casey, Mustafa, Sam, you guys just become code reviewers, and you sort of polish and ship
358 00:43:18.320 ⇒ 00:43:28.259 Uttam Kumaran: Which is much easier, I think, because as you can see, like, there’s so much lack of context on both sides that it’s very, very hard, I think, to ship some of these features.
359 00:43:28.380 ⇒ 00:43:30.569 Uttam Kumaran: Because you might hear 3 lines from me.
360 00:43:30.820 ⇒ 00:43:34.899 Uttam Kumaran: And then we have to do something with it, and that’s not fair to you guys, you know?
361 00:43:35.330 ⇒ 00:43:38.619 Uttam Kumaran: So I think democratizing that is definitely, like, a way to go.
362 00:43:38.960 ⇒ 00:43:39.590 Mustafa Raja: Yeah.
363 00:43:44.730 ⇒ 00:43:45.539 Henry Zhao: Okay, I gotta drop.
364 00:43:45.540 ⇒ 00:43:49.569 Uttam Kumaran: Do you guys have… Yeah, yeah, go ahead, Henry. No worries. Thank you, dude.
365 00:43:49.570 ⇒ 00:43:50.700 Henry Zhao: Yeah, thanks again.
366 00:43:51.630 ⇒ 00:43:56.170 Uttam Kumaran: I guess… thank you. I think Casey or Mustaf, do you guys care about, like.
367 00:43:56.280 ⇒ 00:43:59.800 Uttam Kumaran: Magic patterns versus lovable versus bold, or like…
368 00:43:59.800 ⇒ 00:44:06.410 Mustafa Raja: I… I have tried Bolt and Lovable in the past, like… 8 months ago.
369 00:44:07.290 ⇒ 00:44:10.170 Mustafa Raja: Bolt was good, but was expensive for me.
370 00:44:10.590 ⇒ 00:44:12.639 Mustafa Raja: This Magic Pack.
371 00:44:12.640 ⇒ 00:44:15.410 Uttam Kumaran: You know Bolt, you know Bolt was one of our customers?
372 00:44:15.990 ⇒ 00:44:22.229 Mustafa Raja: Oh, yeah, I did so… I did see it in one of the slides that Hannah made, and I was surprised.
373 00:44:23.110 ⇒ 00:44:26.600 Uttam Kumaran: Friend runs… runs their… Growth.
374 00:44:27.060 ⇒ 00:44:30.420 Mustafa Raja: That’s super cool. Yeah, yeah. So, I didn’t…
375 00:44:30.420 ⇒ 00:44:31.760 Uttam Kumaran: Get some for credits, maybe.
376 00:44:31.760 ⇒ 00:44:48.530 Mustafa Raja: Yeah, that’ll be good. I just felt that it was expensive. It used up a lot of credits really quick. I don’t know how Magic Pattern does this, or how the pricing looks like.
377 00:44:48.660 ⇒ 00:44:57.740 Mustafa Raja: But, Lovable wasn’t able to do much for me. Bold was good, but expensive. And magic patterns do look good.
378 00:44:59.180 ⇒ 00:44:59.850 Uttam Kumaran: Okay.
379 00:45:00.470 ⇒ 00:45:00.950 Mustafa Raja: Yeah.
380 00:45:01.290 ⇒ 00:45:07.560 Casie Aviles: Yeah, we had to try these tools in particular, but I’m curious about magic patterns.
381 00:45:08.330 ⇒ 00:45:20.070 Mustafa Raja: Yeah, and I use those early 8 months or so ago, so I guess I’ll have to use them again to be able to make a decision on that.
382 00:45:22.960 ⇒ 00:45:23.650 Uttam Kumaran: Okay.
383 00:45:25.720 ⇒ 00:45:29.190 Mustafa Raja: Yeah, I’ll give it a go, and I’ll inform.
384 00:45:32.700 ⇒ 00:45:33.260 Uttam Kumaran: Okay.
385 00:45:38.210 ⇒ 00:45:43.319 Uttam Kumaran: So yeah, I’ve heard of lovable… I’ve heard of magic patterns, I think we should test lovable…
386 00:45:43.460 ⇒ 00:45:48.649 Uttam Kumaran: I don’t know if there’s anything else, like, maybe I’ll ask Perplexity what other options there are.
387 00:45:50.690 ⇒ 00:45:52.509 Mustafa Raja: Perplexity is really good.
388 00:45:53.900 ⇒ 00:45:55.760 Uttam Kumaran: Yeah, did you guys get the pro version?
389 00:45:56.460 ⇒ 00:45:57.300 Mustafa Raja: Nope.
390 00:45:57.610 ⇒ 00:45:58.209 Mustafa Raja: I didn’t know.
391 00:45:58.210 ⇒ 00:46:00.600 Uttam Kumaran: If you hook up your PayPal to it, you can get a free program.
392 00:46:01.620 ⇒ 00:46:04.339 Mustafa Raja: Pakistan does not have people.
393 00:46:04.630 ⇒ 00:46:05.930 Uttam Kumaran: Oh, really? Okay.
394 00:46:05.930 ⇒ 00:46:06.310 Mustafa Raja: Yeah.
395 00:46:06.310 ⇒ 00:46:12.130 Uttam Kumaran: Let me know, I can get it for you. I mean, I just… I use it… I don’t Google search at all, ever, anymore, really.
396 00:46:13.460 ⇒ 00:46:14.080 Mustafa Raja: That’s right.
397 00:46:14.080 ⇒ 00:46:15.550 Uttam Kumaran: purely use perplexity.
398 00:46:34.360 ⇒ 00:46:37.339 Uttam Kumaran: UI… U… UI isard?
399 00:46:39.390 ⇒ 00:46:40.880 Uttam Kumaran: OV0?
400 00:46:41.490 ⇒ 00:46:43.630 Mustafa Raja: Oh, Vigil is super nice, have you tried it?
401 00:46:44.240 ⇒ 00:46:45.100 Uttam Kumaran: No.
402 00:46:45.100 ⇒ 00:47:00.860 Mustafa Raja: Yeah, V0 is super nice, you should. I have, so, I did Bolt, I did, Livable, didn’t like them. I stumbled upon V0, and I did stuck with it for a long time. I don’t use it anymore, but I used to use it a lot.
403 00:47:02.740 ⇒ 00:47:05.169 Mustafa Raja: And I have only good things to say about it.
404 00:47:06.680 ⇒ 00:47:08.100 Uttam Kumaran: Oh, interesting.
405 00:47:08.280 ⇒ 00:47:20.330 Mustafa Raja: Yeah. Part of the reason is it uses React to build up components, and then it uses only one UI library, that is ShadCN, to build components.
406 00:47:20.330 ⇒ 00:47:21.110 Uttam Kumaran: Oh, okay.
407 00:47:21.110 ⇒ 00:47:26.430 Mustafa Raja: And that is very flexible. Currently, what we are using is we are using MUI,
408 00:47:26.760 ⇒ 00:47:28.910 Mustafa Raja: Oh, yeah, this is, yeah, Sher Xian.
409 00:47:29.630 ⇒ 00:47:38.760 Mustafa Raja: We are using MUI, and if any of these tools, we have could use MUI, that… that might… that might be… that might be a good…
410 00:47:39.000 ⇒ 00:47:41.080 Mustafa Raja: Plus point? I don’t know.
411 00:47:43.740 ⇒ 00:47:45.340 Uttam Kumaran: Interesting, okay.
412 00:47:46.830 ⇒ 00:47:58.789 Uttam Kumaran: I mean, maybe we should just try all these out and see. I don’t mind using V0. I guess… I think they’re all gonna be decent, but how fast can you go from what it designs to our…
413 00:47:58.980 ⇒ 00:48:00.650 Uttam Kumaran: to cursor, basically.
414 00:48:00.800 ⇒ 00:48:01.250 Mustafa Raja: Yeah.
415 00:48:01.250 ⇒ 00:48:01.889 Uttam Kumaran: you know.
416 00:48:02.220 ⇒ 00:48:03.010 Mustafa Raja: Yeah.
417 00:48:03.010 ⇒ 00:48:04.510 Uttam Kumaran: That’s the thing to figure out.
418 00:48:05.500 ⇒ 00:48:06.180 Mustafa Raja: Yeah.
419 00:48:10.420 ⇒ 00:48:15.170 Uttam Kumaran: Because I still think you need the design step. Like, I don’t think you can just tell cursor and text.
420 00:48:15.740 ⇒ 00:48:16.909 Mustafa Raja: Yeah, I agree with that.
421 00:48:16.910 ⇒ 00:48:18.670 Uttam Kumaran: I think the design step is important.
422 00:48:22.460 ⇒ 00:48:24.109 Mustafa Raja: Casey, have you used B0?
423 00:48:25.270 ⇒ 00:48:27.909 Casie Aviles: No, but yeah, with, with your,
424 00:48:30.360 ⇒ 00:48:33.450 Casie Aviles: Recommendation, sounds like something I should try.
425 00:48:33.800 ⇒ 00:48:34.540 Mustafa Raja: Yeah.
426 00:48:34.830 ⇒ 00:48:38.679 Mustafa Raja: I don’t know how it will compare with magic patterns now, because…
427 00:48:39.140 ⇒ 00:48:41.820 Mustafa Raja: I haven’t used it for quite a while now.
428 00:48:43.150 ⇒ 00:48:46.940 Casie Aviles: I’ve only ever built with Courser directly, but yeah.
429 00:48:47.260 ⇒ 00:48:48.080 Mustafa Raja: Yeah.
430 00:48:51.910 ⇒ 00:48:53.860 Mustafa Raja: I used to do mock-ups with it.
431 00:48:57.460 ⇒ 00:49:00.010 Mustafa Raja: When I used to do mobile applications.
432 00:49:00.490 ⇒ 00:49:01.520 Uttam Kumaran: Oh, really?
433 00:49:01.520 ⇒ 00:49:02.130 Mustafa Raja: Yeah.
434 00:49:12.870 ⇒ 00:49:18.450 Uttam Kumaran: Okay, cool, that’s kind of all I had, so I think that’s, like, something we maybe want to try to do, but…
435 00:49:18.740 ⇒ 00:49:21.969 Uttam Kumaran: Let’s, let’s talk… I’ll talk… I’ll talk to it about,
436 00:49:22.410 ⇒ 00:49:27.600 Uttam Kumaran: I’ll talk to it… to Sam about it tomorrow, and see how we can start to enable more of this.
437 00:49:28.000 ⇒ 00:49:29.160 Uttam Kumaran: That’d be nice.
438 00:49:29.420 ⇒ 00:49:30.020 Mustafa Raja: Okay.
439 00:49:30.830 ⇒ 00:49:33.870 Uttam Kumaran: Also, do you guys see cursor has planning mode now?
440 00:49:35.590 ⇒ 00:49:37.840 Mustafa Raja: I did see that you had agents or something.
441 00:49:38.920 ⇒ 00:49:40.980 Uttam Kumaran: Yeah, let me, let me see…
442 00:49:41.800 ⇒ 00:49:44.189 Uttam Kumaran: I think, you know, Windsurf had plan mode.
443 00:49:44.420 ⇒ 00:49:46.730 Uttam Kumaran: But I think the cursor just added it.
444 00:49:49.730 ⇒ 00:49:51.810 Uttam Kumaran: Let’s see what… let’s see what it pulls up.
445 00:50:07.760 ⇒ 00:50:13.420 Mustafa Raja: I wonder if Cursa would… I wonder if Cursa would be able to analyze if.
446 00:50:13.420 ⇒ 00:50:13.760 Uttam Kumaran: plans.
447 00:50:14.480 ⇒ 00:50:16.980 Mustafa Raja: Yeah, oh yeah, this is just added, yeah.
448 00:50:17.530 ⇒ 00:50:35.349 Mustafa Raja: I wondered if Kesha would be able to analyze if, any of the features that a user would want would need some sort of database design, because the suggestions that Hannah had did require some superbase, tables.
449 00:50:36.180 ⇒ 00:50:42.059 Uttam Kumaran: Oh, yeah, that’s also the thing, is, like, I want… I was trying to see, like, how I can… how we can,
450 00:50:42.970 ⇒ 00:50:45.250 Uttam Kumaran: have Kirscher have… have avail… have…
451 00:50:45.660 ⇒ 00:50:49.360 Uttam Kumaran: give Kirscher more availability of Supabase, you know?
452 00:50:49.940 ⇒ 00:50:50.920 Mustafa Raja: Hmm…
453 00:50:51.610 ⇒ 00:50:54.939 Uttam Kumaran: And can I tell you, like, this document, I used Codex to write.
454 00:50:55.140 ⇒ 00:50:55.920 Mustafa Raja: Oh, the…
455 00:50:56.670 ⇒ 00:51:01.030 Uttam Kumaran: And it took 3 minutes, and thought about the whole plan and wrote it.
456 00:51:02.440 ⇒ 00:51:08.340 Uttam Kumaran: It, like, set up the environment, It, like, looks… Through the entire file.
457 00:51:09.280 ⇒ 00:51:13.680 Uttam Kumaran: So, you guys should try Codex, too. It’s… well, it comes with all of our pro accounts, so…
458 00:51:15.940 ⇒ 00:51:16.360 Mustafa Raja: Yeah.
459 00:51:16.360 ⇒ 00:51:19.460 Uttam Kumaran: And then, look, there’s two bugs on your PR.
460 00:51:19.460 ⇒ 00:51:24.209 Mustafa Raja: Yeah, I’ll look into that.
461 00:51:31.320 ⇒ 00:51:34.980 Casie Aviles: Oh, so this can… this can… Do pull requests…
462 00:51:35.890 ⇒ 00:51:39.349 Uttam Kumaran: Yeah, so I… I went to,
463 00:51:43.780 ⇒ 00:51:46.950 Casie Aviles: And we can just connect it to our repo. Okay, cool.
464 00:51:47.280 ⇒ 00:51:56.169 Uttam Kumaran: Yeah, I just did… I just did add Codex. I know we were trying all of those, I think we’re gonna either use Cursor, BugBot, or… or… or, Codex.
465 00:52:03.080 ⇒ 00:52:04.320 Mustafa Raja: Codex looks good.
466 00:52:24.140 ⇒ 00:52:27.029 Uttam Kumaran: Okay, cool. That’s all I had, guys, so…
467 00:52:27.610 ⇒ 00:52:30.409 Uttam Kumaran: Yeah, I think we’ll try to work on that, and then I’m gonna try to…
468 00:52:30.790 ⇒ 00:52:34.889 Uttam Kumaran: ship one or two things tonight and see if I can get something in a couple branches.
469 00:52:35.520 ⇒ 00:52:37.530 Uttam Kumaran: For y’all to take a look at.
470 00:52:38.630 ⇒ 00:52:39.270 Mustafa Raja: Okay.
471 00:52:39.620 ⇒ 00:52:40.550 Mustafa Raja: Thank you.
472 00:52:41.390 ⇒ 00:52:43.749 Uttam Kumaran: Alright, thanks guys. Talk to you soon.
473 00:52:44.010 ⇒ 00:52:44.800 Casie Aviles: Thank you.
474 00:52:44.800 ⇒ 00:52:45.450 Mustafa Raja: Thank you, bye.
475 00:52:45.450 ⇒ 00:52:46.080 Uttam Kumaran: Bye.