Meeting Title: New workflow share + Latest Model Access Date: 2026-04-15 Meeting participants: Mustafa Raja, Casie Aviles, Samuel Roberts, Uttam Kumaran, Pranav Narahari, Awaish Kumar
WEBVTT
1 00:00:28.070 ⇒ 00:00:29.230 Samuel Roberts: Hello.
2 00:00:32.140 ⇒ 00:00:32.790 Mustafa Raja: Hey.
3 00:00:33.910 ⇒ 00:00:34.360 Casie Aviles: Hmm.
4 00:00:34.360 ⇒ 00:00:37.039 Samuel Roberts: Do I sound okay? I changed my microphone settings.
5 00:00:37.940 ⇒ 00:00:39.420 Mustafa Raja: Yeah, you sound good.
6 00:00:39.630 ⇒ 00:00:40.450 Samuel Roberts: Okay, cool.
7 00:00:43.660 ⇒ 00:00:49.620 Samuel Roberts: I’m… also having to clean off my screen, because I… I usually have my desk set up upstairs.
8 00:00:49.730 ⇒ 00:00:50.800 Samuel Roberts: And
9 00:00:51.170 ⇒ 00:00:56.320 Samuel Roberts: I don’t… I haven’t used a laptop out in this, like, light in a while, so my screen was nasty.
10 00:00:56.800 ⇒ 00:00:57.450 Mustafa Raja: No.
11 00:00:57.960 ⇒ 00:01:03.870 Samuel Roberts: But… I’m enjoying the little time away from those big monitors sometimes, too. Hey, Oh, Tom.
12 00:01:04.810 ⇒ 00:01:05.750 Uttam Kumaran: Hey guys.
13 00:01:06.250 ⇒ 00:01:06.940 Mustafa Raja: Hmm.
14 00:01:07.930 ⇒ 00:01:09.940 Samuel Roberts: So excited to hear what’s going on here.
15 00:01:10.540 ⇒ 00:01:12.019 Uttam Kumaran: Yeah, so I’ve…
16 00:01:12.970 ⇒ 00:01:19.899 Uttam Kumaran: I have… I just was getting kind of, like, tired about a couple things for Cursor. One is, like, we’re starting to pay a lot of money.
17 00:01:20.120 ⇒ 00:01:26.440 Uttam Kumaran: And it’s, basically the way Cursor works is it’s like a tax on top of,
18 00:01:26.650 ⇒ 00:01:43.090 Uttam Kumaran: the existing model costs. So on one lens, it’s actually better for us to just go direct, but also, like, I kind of… now that we have the platform and the knowledge set up in a certain way, you actually really don’t need to, like.
19 00:01:43.970 ⇒ 00:01:52.840 Uttam Kumaran: you don’t necessarily need to, like, see the structure all the time, and in fact, I think for a lot of the whole team’s work, it’s actually…
20 00:01:53.310 ⇒ 00:01:54.420 Uttam Kumaran: the GitHub…
21 00:01:54.660 ⇒ 00:02:03.360 Uttam Kumaran: structure is actually really less relevant. It’s actually more about, like, is the agent actually end-to-end able to, like, solve its own problems?
22 00:02:03.840 ⇒ 00:02:23.339 Uttam Kumaran: And so, like, my workflow to date has been probably what you guys are familiar with, so probably the last 6 months, I’ve been really heavy on, like, cursor. I sort of skipped a lot of, like, cloud code and stuff like that, because I don’t know, I, like, I got it back then, but we had all this, like, platform structure work that, like, I needed, sort of, cursor for, and so, kind of, like.
23 00:02:23.390 ⇒ 00:02:30.480 Uttam Kumaran: what I’m trying to think more about is, like, as a developer at Brainforge, you’re gonna have, like, multiple ways to, like, execute work.
24 00:02:30.700 ⇒ 00:02:34.710 Uttam Kumaran: I think we’re gonna continue to use IDE-related environments, but I’ll…
25 00:02:34.840 ⇒ 00:02:37.030 Uttam Kumaran: I’ll describe those more like scalpels.
26 00:02:37.160 ⇒ 00:02:53.500 Uttam Kumaran: like, you need to, like, look at a specific issue, you need to run something. This is sort of more of, like, a chat back and forth. I’ll also… I’m also gonna share today that I’ve been using, like, OpenCode, which is, like, an open source version of Claude Code, and it’s…
27 00:02:53.510 ⇒ 00:02:59.069 Uttam Kumaran: sick. It works really well, and I’ll kind of show you why I feel like it’s actually been a lot…
28 00:02:59.180 ⇒ 00:03:18.760 Uttam Kumaran: it’s like, I actually have gotten, like, much more done, faster than if I was to do the same thing in Cursor. And, another reason is, like, we have a bunch of Azure credits that I need to spend, actually. So, I can’t do that with Cursor, and so actually, like, I’m able to unlock for this group.
29 00:03:18.870 ⇒ 00:03:22.019 Uttam Kumaran: Like, basically 5.4, 5.4 mini.
30 00:03:22.120 ⇒ 00:03:37.900 Uttam Kumaran: Which hopefully should also be helpful to execute work. And then the last surface that I’ll talk about is I’m actually, like, I’m just, like, wrapping it up, but I’ve gotten, I’m basically gonna allow us to start to assign work to a Brainforge platform agent.
31 00:03:37.920 ⇒ 00:03:56.790 Uttam Kumaran: Similarly, in linear, you should be seeing that you can assign work to cursor and to codex. Well, I’ve gotten, like, open codes set up where you can actually assign it to the platform, and I’m just, like, dealing with some, like, linear auth callback thing, but, like, that should be out today. And the reason…
32 00:03:56.840 ⇒ 00:04:08.730 Uttam Kumaran: So I’ll kind of… I’m gonna walk through, like, these three examples, and then I’ll just, like, leave it open for questions, and then all the, instruction on how to set up open code, CLI,
33 00:04:08.740 ⇒ 00:04:17.380 Uttam Kumaran: It’s all in the platform, so you can just go into Kirscher and ask, like, help me set this up. But let me just share my screen. This is gonna be, like, really…
34 00:04:17.470 ⇒ 00:04:25.130 Uttam Kumaran: like, a raw, unprepared meeting, so just bear with me, but I kind of, like, want… you guys are all, like, technical, so I just want to, like, jump into it.
35 00:04:26.310 ⇒ 00:04:43.060 Uttam Kumaran: So, what I have now on my screen is sort of, like, several open code CLI views that I’m executing, but you can see that I’m also still in linear… still in cursor vanilla, right, for stuff. Like, for example, I’m working on this to-do briefing.
36 00:04:43.360 ⇒ 00:04:52.339 Uttam Kumaran: like, skill right now. Like, it’s kind of nice for me to see it here, like, I don’t need to open up another CLI to do that, blah blah blah.
37 00:04:52.590 ⇒ 00:04:53.910 Uttam Kumaran: So, that’s…
38 00:04:55.290 ⇒ 00:05:00.430 Uttam Kumaran: Cannot use this meeting transcript, please, as the host to upgrade? That was just a note, I was just…
39 00:05:00.430 ⇒ 00:05:04.170 Samuel Roberts: That’s related, I just was putting it there, so I… I didn’t want to interrupt you, sorry.
40 00:05:04.270 ⇒ 00:05:19.760 Uttam Kumaran: No, you’re good again. So… so on this right side, so here’s a classic task that everybody may be used to. I’m working on a skill. I’m working on updating the daily to-do briefing skill. I like this skill because everybody, you should be able to go in and be like, what am I… what do I have to do today?
41 00:05:20.120 ⇒ 00:05:39.410 Uttam Kumaran: And, like, I think this is gonna be great. So this is, like, more of a scalpel task, where I’m, like, the skill exists, I need to just modify it in a little bit of ways, sure, I’m gonna use cursor. But here’s an example of something that I’m working on, right now, which is basically, like, let me try to show you guys, like, what this is.
42 00:05:41.970 ⇒ 00:05:50.209 Uttam Kumaran: So, I’m working on… in this PR, I’m actually working on getting, Let’s see…
43 00:05:52.300 ⇒ 00:06:10.759 Uttam Kumaran: Oh, yeah, so I’m doing an end-to-end deployment of OpenPanel. OpenPanel is a product analytics platform that I’m gonna use to basically measure how folks at Brainford use Platform, use our new work product, how they’re gonna use OpenCode, and so I basically, I’m having
44 00:06:11.210 ⇒ 00:06:29.180 Uttam Kumaran: a single OpenCode instance running since, like, yesterday around 6, where I’m, like, sort of steering it to actually do the end-to-end deployment. And one thing I love about OpenCode CLI is that, it’s, like, actually, like, more persistent. Like, I think Cursor…
45 00:06:29.240 ⇒ 00:06:43.359 Uttam Kumaran: sometimes sucks, because it keeps stopping, and it doesn’t, like, it’s not resourceful. It doesn’t look that, oh yeah, there is context on how to connect a railway, I could probably just connect a railway. Like, the reasoning isn’t good enough, I don’t know what it is.
46 00:06:43.360 ⇒ 00:06:54.270 Uttam Kumaran: So I’m finding that it’s actually much better in terms of actually being able to take a task and drive it end-to-end, and then pause only where there’s actually a real need for
47 00:06:54.490 ⇒ 00:07:04.179 Uttam Kumaran: like, support. So kind of, like, the way you can bring this up, you know, once you guys have this set up, is I… I have… I have a command on my thing called OC,
48 00:07:04.580 ⇒ 00:07:12.859 Uttam Kumaran: legacy. The reason it’s open code legacy is because we’re using a legacy… we’re using our legacy deployment for Azure.
49 00:07:12.920 ⇒ 00:07:31.190 Uttam Kumaran: Because there’s a certain, like, responses API that are… that’s only allowed on a few of the models on that deployment. But, like, again, a lot of this is already in there, is in… is in the platform and documentation. So really, like, what I… what I’m thinking about
50 00:07:31.260 ⇒ 00:07:34.750 Uttam Kumaran: Is… you could actually open this, and you can basically start to code.
51 00:07:34.880 ⇒ 00:07:41.059 Uttam Kumaran: But you could also… you can trigger anything, like, you can trigger all the skills that are in the platform are all here.
52 00:07:41.160 ⇒ 00:07:50.769 Uttam Kumaran: And basically, I think this is a much better surface for development work that I found in the last, like, 5-6 days versus Cursor.
53 00:07:50.890 ⇒ 00:08:00.490 Uttam Kumaran: However, I’m still, like, running it within a cursor IDE, which is great, because you can sort of stack a couple things. And I’m also, also connected to all the, like, MCPs.
54 00:08:00.940 ⇒ 00:08:20.880 Uttam Kumaran: So I have, like, a couple of the MCPs that I like, Supabase, some of these I just don’t have… there’s no CLI for, or some of these that MCP is just better. So I still have my, like, MCPs here, and then I’m actually just, like, executing, so I can say, hey, take on this linear ticket and execute it, and I just feel like the odds of success have been way higher.
55 00:08:20.890 ⇒ 00:08:35.090 Uttam Kumaran: Especially for, like, full end-to-end, like, full-stack type work. Additionally, as you can see, it, like, doesn’t stop. Like, one of the things that sucks about Versor is that you can have multiple chats, but it stops, and so I’m finding myself
56 00:08:35.150 ⇒ 00:08:46.479 Uttam Kumaran: going between, like, 5 of them and, like, babysitting. I think this has actually been a lot better at, like, driving things to the finish line and using all the tools you give it.
57 00:08:46.620 ⇒ 00:08:54.959 Uttam Kumaran: Whether for me it’s, like, Chrome DevTools, whether it’s, like, Railway CLI, whether it’s access to 1Password through the CLI,
58 00:08:55.020 ⇒ 00:08:58.390 Uttam Kumaran: So I’ve just found this to be a lot more effective.
59 00:08:58.430 ⇒ 00:09:10.399 Uttam Kumaran: So, for example, you could basically say, give it an end-to-end task, and I think still the principle of, like, work on the plan first, and then work on the,
60 00:09:10.400 ⇒ 00:09:19.010 Uttam Kumaran: like, work on the plan first, give it… give it the plan, and then, like, have it execute is still, like, the best sort of way to interact. But,
61 00:09:19.200 ⇒ 00:09:34.650 Uttam Kumaran: this is now, like, available for everyone to use. There’s also an open code desktop, which is basically, like, an alternative to cursor. I don’t have, like, it all working. You’re free to go try that. But that’s probably where…
62 00:09:34.840 ⇒ 00:09:39.100 Uttam Kumaran: We will start to shift a lot of stuff internally to, is…
63 00:09:39.230 ⇒ 00:09:52.799 Uttam Kumaran: people at BrainForge can bring whatever else, but I’m gonna… I’m gonna start to configure OpenCode, OpenCode Desktop, the Brainforce platform, background agents, to just be really tuned to us.
64 00:09:52.950 ⇒ 00:10:02.569 Uttam Kumaran: And so it’s… it’s just gonna be extremely effective at operating in our environment, versus, like, Purser’s agent needs to support, like, every single person that pays them.
65 00:10:02.700 ⇒ 00:10:14.609 Uttam Kumaran: We don’t need all that. We need, actually, the specifics. So for me to push open code to suggest playbooks, for me to push open code to use certain skills, to think about our clients, I’m able to actually do. So.
66 00:10:15.330 ⇒ 00:10:19.739 Uttam Kumaran: I’ll kind of pause there. Any questions?
67 00:10:19.860 ⇒ 00:10:21.240 Uttam Kumaran: Or any thoughts.
68 00:10:23.730 ⇒ 00:10:31.730 Samuel Roberts: So for this right now, you have a terminal down here, and then up here on the right, is that another terminal?
69 00:10:32.480 ⇒ 00:10:36.909 Samuel Roberts: For open code, so that’s how you’re running multiple ones, you just have several terminal windows open in cursor?
70 00:10:38.210 ⇒ 00:10:51.079 Uttam Kumaran: Yes, so I basically… so what you’re gonna find is… is the amount of window… the amount of open code CLI you can have is just gonna be limited to your, sort of, RAM. So if you look at, like, my activity monitor.
71 00:10:51.170 ⇒ 00:11:05.529 Uttam Kumaran: So, okay, so actually, let me tell you even more of, like, why I… I’m, like, actually kind of pissed at Cursor, is, like, I could… typically, I could only have, like, 3 or 4 of these in addition to Cursor up before my… my, like, 24GB RAM
72 00:11:05.720 ⇒ 00:11:12.069 Uttam Kumaran: Mac Mini fries. My… I have an 8GB… MacBook Air, that’s even worse.
73 00:11:12.250 ⇒ 00:11:25.720 Uttam Kumaran: And something got nuked with Cursor in the past, like, 6 weeks that sucks. Like, it now takes a lot of RAM, and I think they nuked Auto. Like, auto sucks. And so I kinda am tired of it.
74 00:11:25.980 ⇒ 00:11:43.340 Uttam Kumaran: I… I want to get… make… I’m sure you guys are also, like, noticing a little bit of that, so this is gonna be the solve. So yeah, what I do is I just create a new terminal window, and then I just drag it to the corner, and if you… if you close your… some of your windows… I was sustaining 8 of these yesterday on, like, a…
75 00:11:43.340 ⇒ 00:11:49.659 Uttam Kumaran: 24 gig machine. I actually think Cursor, had I just run this in separate CLI windows.
76 00:11:49.790 ⇒ 00:11:52.720 Uttam Kumaran: Or use Conductor or something.
77 00:11:52.820 ⇒ 00:12:02.989 Uttam Kumaran: you may have gone… been able to do a lot more. For me, I’m… I’m limited to sort of, like, my context, also my, like, mental context, so typically I have, like, 4 to 6 of these running.
78 00:12:04.090 ⇒ 00:12:04.910 Uttam Kumaran: Yeah.
79 00:12:04.910 ⇒ 00:12:18.430 Samuel Roberts: If anyone’s used, CMUX, I’ve been playing around with that as a terminal. It has… does kind of multiple windows very easily, kind of like a Tmux thing, but a little more graphical, and it surfaces, like, when things need you and things like that, kind of like Conductor, but just pure terminal.
80 00:12:18.540 ⇒ 00:12:20.720 Samuel Roberts: If anyone wants to check that out.
81 00:12:20.990 ⇒ 00:12:22.609 Samuel Roberts: That may be how I start running this.
82 00:12:22.770 ⇒ 00:12:25.190 Samuel Roberts: Because cursor is eating a ton of my memory.
83 00:12:29.710 ⇒ 00:12:30.660 Samuel Roberts: Cool.
84 00:12:31.870 ⇒ 00:12:33.649 Uttam Kumaran: What… what are the thoughts? Yeah.
85 00:12:34.030 ⇒ 00:12:38.579 Samuel Roberts: I was gonna say, have you… so you haven’t played with the GUI yet for OpenCode at all?
86 00:12:38.710 ⇒ 00:12:43.989 Samuel Roberts: But we can try to get that running, because I’ve been using the agent mode in cursor a lot more, which is much more, like…
87 00:12:44.450 ⇒ 00:12:47.129 Samuel Roberts: you know, less IDE, more just agent.
88 00:12:48.430 ⇒ 00:12:58.239 Uttam Kumaran: Yeah, I mean, you can… so you can run Open Code Desktop. Actually, like, there are instructions on how to do it. It’s just a little finicky, and I…
89 00:12:58.460 ⇒ 00:13:03.040 Uttam Kumaran: Yeah, I’m just getting the basics, because I need to use it to ship a bunch of shit, so I was like.
90 00:13:03.040 ⇒ 00:13:03.590 Samuel Roberts: Right, right, right.
91 00:13:03.590 ⇒ 00:13:21.739 Uttam Kumaran: much better, but yeah, you can get the OpenCode desktop working pretty easily. Like, everybody on this call should be able to get that going. I think the biggest thing I want everybody to try to test today is that you can get into Azure. Like, you can use 1Password to get the Azure credentials, and ideally, you should just see this.
92 00:13:21.740 ⇒ 00:13:31.109 Uttam Kumaran: what are some things that you’ll know if you’re wrong? If the AI responds to you, like, I cannot support you, that request timeout, some other issues.
93 00:13:31.140 ⇒ 00:13:45.539 Uttam Kumaran: Like, you’re probably connected to the wrong Azure, or there’s some hookup that’s wrong. OpenCode also comes out with a lot of free models, which is sick, and they’re all… they’re all, like, better than auto. Like, so there are these, like,
94 00:13:46.030 ⇒ 00:13:51.319 Uttam Kumaran: They have, like, Big Pickle, Nemo Tron, and then they have, like, another one that’s, like,
95 00:13:52.550 ⇒ 00:13:58.759 Uttam Kumaran: I forgot what the other one was, like, Like, the Zen…
96 00:13:59.030 ⇒ 00:14:02.499 Uttam Kumaran: They have a couple… they have a couple ones that… that are, like, free.
97 00:14:02.600 ⇒ 00:14:11.249 Uttam Kumaran: That are also pretty good. And so, I also… I’m kind of interested in this group trying out some of the models that come bundled, and then I can pay for that, because
98 00:14:11.440 ⇒ 00:14:17.929 Uttam Kumaran: those are, like, 100 bucks. Like, to give you guys a sense, we’re, like, we’re kind of running at, like, 3 grand in cursor costs a month.
99 00:14:18.280 ⇒ 00:14:23.949 Uttam Kumaran: Like, 500 of that is, like, me, but, like, it’s kind of ridiculous.
100 00:14:24.110 ⇒ 00:14:25.689 Uttam Kumaran: And we’re… and, like.
101 00:14:25.980 ⇒ 00:14:31.119 Uttam Kumaran: And I think you can… so I’m thinking about moving us to, like, a mix of open code desktop.
102 00:14:31.400 ⇒ 00:14:44.109 Uttam Kumaran: We’re gonna come out with this, like, Claude Coworker clone also, so the people that are mainly doing, like, planning and stuff can use that. And then I’m also gonna release the background agents. It’s not only gonna be cheap… I’m not really… I don’t really want…
103 00:14:44.490 ⇒ 00:14:54.439 Uttam Kumaran: this to be, like, oh, it’s cheaper. This is… this has actually, like, been so much better, also, in the past, like, 3 days for me. Like, it actually is executing work end-to-end.
104 00:14:57.980 ⇒ 00:15:01.660 Awaish Kumar: When you say it’s actually… like…
105 00:15:01.780 ⇒ 00:15:07.020 Awaish Kumar: Working end-to-end, that means, like, it’s not stopping for such, like, getting,
106 00:15:07.290 ⇒ 00:15:11.020 Awaish Kumar: Approval from request to search on the web, and things like that.
107 00:15:12.560 ⇒ 00:15:22.329 Uttam Kumaran: Yeah, well, this is where, like, I think if you’re… if you want to search on the web, you should use the EXA MCP, so there… we have some documentation on that.
108 00:15:22.720 ⇒ 00:15:29.349 Awaish Kumar: So, like, the question itself goes on, like, if we have a prompt and… Sometimes it goes on to…
109 00:15:29.840 ⇒ 00:15:30.550 Awaish Kumar: browse.
110 00:15:30.550 ⇒ 00:15:39.189 Uttam Kumaran: But I… yeah, so I always say the browser is good, but still, it’s not, like, great, like, it takes a while to load, so I’ve been using Chrome
111 00:15:39.470 ⇒ 00:15:44.330 Uttam Kumaran: Devtools, MCP, and EXA.
112 00:15:44.950 ⇒ 00:15:53.459 Uttam Kumaran: XI is more for web search, Chrome DevTools, what it’ll do is it’ll bring up Chrome, and then it can interact. So, for example, I was having it
113 00:15:53.750 ⇒ 00:16:06.220 Uttam Kumaran: literally click UI buttons to… for my Slack assistant stuff yesterday, and, like… so, like, anything… so, like, if you… I think if you have questions on mechanics, great, also position to ask me, because…
114 00:16:06.380 ⇒ 00:16:14.659 Uttam Kumaran: I’m sort of, like, giving… and this is what we’re calling harnesses, right? So, like, I’m giving the AI as much ability to do all the things
115 00:16:15.330 ⇒ 00:16:18.120 Uttam Kumaran: you know, via CLI as possible.
116 00:16:18.650 ⇒ 00:16:19.690 Uttam Kumaran: Yeah.
117 00:16:21.510 ⇒ 00:16:28.179 Samuel Roberts: Oh, wait, are you wondering about, like, allowing certain tool calls versus just, like, when cursor stops and it thinks it’s done kind of thing?
118 00:16:28.920 ⇒ 00:16:34.950 Awaish Kumar: Yeah, like, when… When are the things when the customer stops, like, asking for approval and things like that?
119 00:16:34.950 ⇒ 00:16:35.660 Samuel Roberts: Yeah.
120 00:16:35.660 ⇒ 00:16:40.829 Awaish Kumar: Is it… is it bypassing that? Or, like, when you say end-to-end, like, what that exactly means?
121 00:16:45.030 ⇒ 00:16:52.919 Uttam Kumaran: I’m talking about end-to-end, like, meaning something can search the web, get the information, and then execute, deploy to railway.
122 00:16:53.200 ⇒ 00:17:00.969 Uttam Kumaran: open the PR, see the PRCI, log into Snowflake, execute dbt code, like, everything.
123 00:17:01.830 ⇒ 00:17:10.359 Uttam Kumaran: I want you to be able to basically be like, here’s a ticket, like, execute, and then as it gets stuck, or, like, it goes too far, is when we should…
124 00:17:10.720 ⇒ 00:17:14.779 Uttam Kumaran: We should, like, you know, fix, you know?
125 00:17:14.780 ⇒ 00:17:19.560 Samuel Roberts: Are you… Utam, are you running in, like, a YOLO mode where it’s not prompting for any tool calls, or have you…
126 00:17:19.560 ⇒ 00:17:23.500 Uttam Kumaran: Yeah, I’m not interested… yeah, I’m not interested in it, like, prompting at all. Yeah.
127 00:17:23.500 ⇒ 00:17:37.880 Samuel Roberts: Yeah, I think that’s what Awas, yeah, Oasis, there’s a way to set it so that it either asks you every time for everything, or you can allow certain ones, or you can just allow everything, and I think Utam’s got open code running with everything allowed. That’s why he’s able to go fully end-to-end, and it doesn’t stop in the middle.
128 00:17:38.470 ⇒ 00:17:50.939 Uttam Kumaran: Yeah. So, so the open code… the open code config in there is configured this way. So that’s why, like, you could install open code on your own, but I would suggest you just use my configuration. It has all these settings.
129 00:17:51.310 ⇒ 00:17:51.850 Samuel Roberts: Okay.
130 00:17:52.250 ⇒ 00:17:56.189 Samuel Roberts: And that configuration has the stuff for 1Password to pull from…
131 00:17:56.300 ⇒ 00:17:58.860 Samuel Roberts: Or to pull from 1Password to get the Azure keys and everything?
132 00:17:59.360 ⇒ 00:18:00.000 Uttam Kumaran: Yes.
133 00:18:00.240 ⇒ 00:18:00.800 Samuel Roberts: Okay.
134 00:18:01.300 ⇒ 00:18:04.509 Uttam Kumaran: Yeah. And, and it should, it should, it should,
135 00:18:04.780 ⇒ 00:18:08.900 Uttam Kumaran: In the CLI, it’s hard-coded to do the OP pull.
136 00:18:09.250 ⇒ 00:18:10.440 Uttam Kumaran: From, from.
137 00:18:10.440 ⇒ 00:18:13.789 Samuel Roberts: Perfect, so as long as we have it, 1Password, which we should. Okay, cool.
138 00:18:14.170 ⇒ 00:18:14.920 Samuel Roberts: Love it.
139 00:18:16.570 ⇒ 00:18:17.750 Samuel Roberts: I’m excited for this.
140 00:18:18.380 ⇒ 00:18:19.540 Pranav Narahari: So…
141 00:18:19.760 ⇒ 00:18:37.859 Pranav Narahari: Kind of on the end-to-end question, how it doesn’t stop when it thinks it’s completed one part of things, or it doesn’t… it feels like it can’t retrieve certain information, so it kind of stops, like, the request there. Have you noticed that it can kind of go into, like, repetitive loops of just, like.
142 00:18:38.110 ⇒ 00:18:39.310 Pranav Narahari: Messing up.
143 00:18:39.440 ⇒ 00:18:44.770 Pranav Narahari: Trying to figure it out, messing up, and are you, like… is now kind of the process of, like…
144 00:18:45.030 ⇒ 00:18:49.929 Pranav Narahari: You see it going into these loops, it’s not gonna make any further progress, and that’s where you step in?
145 00:18:50.410 ⇒ 00:18:59.809 Pranav Narahari: Or have you just felt like end-to-end is just doing things faster, and it’s also able to, like, use the context very effectively?
146 00:19:01.000 ⇒ 00:19:03.870 Uttam Kumaran: I think it’s really good, because, like.
147 00:19:03.970 ⇒ 00:19:19.800 Uttam Kumaran: And I’ll give us and me a lot of credit. The platform has so much shit in it that it’s able to really figure shit out. Like, you’re not, like, raw-dogging Clawed code in, like, an empty repo. There is so much stuff in here about how we do work.
148 00:19:19.990 ⇒ 00:19:37.359 Uttam Kumaran: that, like, it’s able to… it’s able to run. For example, 20 minutes ago, I just… I’m having this issue with this OAuth thing, and I was futzing around in cursor for, like, an hour, and I literally said, like… I just, like, was like, give me the summary of, like, what you’ve done and what needs to be done. Gonna give it to OpenCode.
149 00:19:37.690 ⇒ 00:19:43.059 Uttam Kumaran: And it was able to, like, continue on that, and as you can see, like.
150 00:19:43.350 ⇒ 00:19:51.369 Uttam Kumaran: it’s still running this stuff for OpenPanel, and you can see it’s looking into railway, it’s pushing a PR, it’s running the deployment steps,
151 00:19:51.580 ⇒ 00:19:53.149 Uttam Kumaran: So I’m finding that
152 00:19:53.450 ⇒ 00:20:08.070 Uttam Kumaran: But this is where I would separate knowledge work from, like, development work, you know? But, like, forgetting Pranav, I think you could get… I could… you could basically give it the entire project, and be like, here’s the JSON, here’s what we’re trying to achieve.
153 00:20:08.320 ⇒ 00:20:21.050 Uttam Kumaran: I want you to work step-by-step and check in with me as you get through steps, but, like, run, you know? Yeah. And that way, it does a lot. It’s a little bit slower than Cursor, but, like.
154 00:20:21.100 ⇒ 00:20:31.530 Uttam Kumaran: you can go do other stuff, and then just check in. The other thing is I had it just give me a little ding when it sends me a little, like, little, like, sound ding when it…
155 00:20:31.610 ⇒ 00:20:33.509 Uttam Kumaran: Finishes a chat, so…
156 00:20:33.770 ⇒ 00:20:44.060 Uttam Kumaran: I’m, like, on AirPods or headphones all day, so… that way, like, you have 4 or 5 of these running, the dings come in once every 10-15 minutes, and I just, like, push it forward.
157 00:20:46.460 ⇒ 00:20:59.759 Uttam Kumaran: And what you’re gonna find is that I’m gonna make sure the platform has playbooks and things like that, so the agents are gonna know where to look when they get stuck much better. But instead, I want you guys to focus on the plan, and when you want to be
158 00:21:00.080 ⇒ 00:21:01.620 Uttam Kumaran: looped in, you know?
159 00:21:02.030 ⇒ 00:21:02.410 Samuel Roberts: Yeah.
160 00:21:03.180 ⇒ 00:21:03.640 Pranav Narahari: Project.
161 00:21:03.640 ⇒ 00:21:15.020 Samuel Roberts: On the, on the dings and stuff, if you’re using another tool instead of cursor, like Conductor or CMUX or just a terminal, like, those often have a way to set that up as well, that will prompt you when something is needed.
162 00:21:15.960 ⇒ 00:21:19.560 Samuel Roberts: So that if anyone has any questions about that, Mike, I can help with that, because I’ve done that a bit.
163 00:21:20.470 ⇒ 00:21:23.280 Samuel Roberts: I had another question about… oh, so for…
164 00:21:23.650 ⇒ 00:21:41.020 Samuel Roberts: for working in other repos, like the ABC or the Eden AI repo, in Cursor, you’re kind of able to, like, open things into the workspace so that it still has context of the platform skills. Is OpenCode CLI able to do something similar? Like, if we’re working somewhere else, but, like, there’s a
165 00:21:41.310 ⇒ 00:21:47.089 Samuel Roberts: skill or a playbook we want to try to do that’s in the platform repo, but we’re working in Eden or ABC or something?
166 00:21:48.620 ⇒ 00:21:49.929 Uttam Kumaran: Wait, say that one more time?
167 00:21:50.190 ⇒ 00:21:51.280 Samuel Roberts: So, like, your cursor…
168 00:21:51.280 ⇒ 00:21:52.370 Uttam Kumaran: the skills.
169 00:21:52.370 ⇒ 00:21:53.029 Samuel Roberts: Yeah, like.
170 00:21:53.030 ⇒ 00:21:53.400 Uttam Kumaran: Yeah.
171 00:21:53.400 ⇒ 00:21:59.120 Samuel Roberts: if you’re in one other repo, I can’t necessarily see those skills, and in Cursor, you’re able to open a whole, like, workspace with several different folders, or…
172 00:21:59.120 ⇒ 00:22:09.390 Uttam Kumaran: Yeah, I have… I have found that it hasn’t really struggled, like, as long as your GitHub PAT, as long as you… as long as…
173 00:22:09.530 ⇒ 00:22:12.150 Uttam Kumaran: You have, like, a personal access token, or whatever.
174 00:22:12.150 ⇒ 00:22:13.420 Samuel Roberts: Oh, sure, okay.
175 00:22:13.420 ⇒ 00:22:18.550 Uttam Kumaran: It’s gonna just figure it out, like, you can switch, like, I was working on default.
176 00:22:18.980 ⇒ 00:22:23.750 Uttam Kumaran: And I don’t think where you open the CLI really, like, impacted much for me, because…
177 00:22:23.750 ⇒ 00:22:24.250 Samuel Roberts: Okay.
178 00:22:24.490 ⇒ 00:22:30.570 Uttam Kumaran: it’s just gonna check it out as it needs, I think, or look for it, but it’s an interesting thing.
179 00:22:30.940 ⇒ 00:22:34.440 Uttam Kumaran: Yeah, and again, I’m not… I’m hopeful that we don’t have, like.
180 00:22:34.800 ⇒ 00:22:45.050 Uttam Kumaran: I don’t know at this point if we’re gonna have client-specific skills, so really, I’m sort of centralized to the platform stuff, and then I… anytime I do client work, I have the platform open, and then I have the other repo.
181 00:22:45.220 ⇒ 00:22:46.520 Samuel Roberts: That’s, yeah, that’s how I’m doing.
182 00:22:46.520 ⇒ 00:22:55.650 Uttam Kumaran: But I’m usually… but also, what I do is I start from the ticket. So if it’s a client work, I always, like, I’m like, does this ticket have everything? And then I just hand it the ticket, typically.
183 00:22:55.650 ⇒ 00:22:56.270 Samuel Roberts: Fair enough.
184 00:22:56.610 ⇒ 00:22:59.069 Uttam Kumaran: And then it’ll infer everything from there.
185 00:22:59.510 ⇒ 00:23:01.099 Samuel Roberts: That’s smart. Okay, I like that.
186 00:23:03.560 ⇒ 00:23:04.240 Samuel Roberts: Cool.
187 00:23:10.690 ⇒ 00:23:12.920 Samuel Roberts: I’m excited for this. I can’t wait to play with it.
188 00:23:13.620 ⇒ 00:23:14.590 Pranav Narahari: Yeah, same.
189 00:23:15.450 ⇒ 00:23:32.469 Mustafa Raja: Yes, I set up OpenCode Desktop earlier. I didn’t properly set it up, it was about a month ago, so I’m excited, on setting it up again with this Azure credentials. I believe Sam and I were in a meeting when we tried to set it up.
190 00:23:32.710 ⇒ 00:23:34.660 Samuel Roberts: Yeah, I forgot about that.
191 00:23:35.870 ⇒ 00:23:50.739 Uttam Kumaran: Yeah, try it. I believe I have all the instructions there, I slogged through this. And then, like, the next thing I will share is I also, like, I have this all working, like, I also pushed the PR on how to set up Tailscale and Terminus.
192 00:23:50.750 ⇒ 00:23:59.880 Uttam Kumaran: So that, like, in case you’re away from, like, your home computer, and you, like, need RAM, like, you can log in. And so, I was…
193 00:23:59.900 ⇒ 00:24:01.910 Uttam Kumaran: Doing stuff from my phone.
194 00:24:01.950 ⇒ 00:24:07.760 Uttam Kumaran: like, when I was at this meeting yesterday, and basically, what’s nice is that you can…
195 00:24:07.890 ⇒ 00:24:13.489 Uttam Kumaran: There is a PR already. You basically… Tailscale just sets up, like, a single node
196 00:24:13.570 ⇒ 00:24:31.360 Uttam Kumaran: like, whatever, like, IP that you log into, use your password, and then you can basically just open up… it just opens up a CLI on your desktop. The nice thing about a CLI on my desktop is I can just go in here, and I can hit CTRL-P, and then look at, like, all my sessions.
197 00:24:31.500 ⇒ 00:24:43.080 Uttam Kumaran: And so, you can just scroll through all your active sessions today. And so, like, for example, my MacBook is fried. I think I’ll probably get another one, like, in a few weeks, but
198 00:24:43.190 ⇒ 00:24:55.090 Uttam Kumaran: now that I can just CLI into here, I don’t need the RAM on my machine that much, so that’s another thing for folks that are traveling. I think that’s gonna be helpful, and there’s a remote dev playbook
199 00:24:55.200 ⇒ 00:25:01.169 Uttam Kumaran: That’s in the repo now, on how to do remote development. I don’t have, a Brainforge
200 00:25:01.290 ⇒ 00:25:04.359 Uttam Kumaran: Tailscale team yet.
201 00:25:04.510 ⇒ 00:25:08.690 Uttam Kumaran: because I have to… there’s some, like, issue with the auth, or some admin thing, but…
202 00:25:08.690 ⇒ 00:25:09.530 Samuel Roberts: Okay, yeah.
203 00:25:09.530 ⇒ 00:25:12.090 Uttam Kumaran: I will have that, but there’s… the playbook is there.
204 00:25:13.270 ⇒ 00:25:16.740 Samuel Roberts: What were you using on your phone? Did Tailscale just give you the CLI, or did you need another app?
205 00:25:16.910 ⇒ 00:25:17.660 Samuel Roberts: to, like…
206 00:25:17.660 ⇒ 00:25:19.280 Uttam Kumaran: I use Terminus.
207 00:25:19.600 ⇒ 00:25:21.919 Samuel Roberts: Oh, Terminator, okay, I knew that sounded familiar.
208 00:25:21.920 ⇒ 00:25:22.310 Uttam Kumaran: Yeah.
209 00:25:22.310 ⇒ 00:25:24.450 Samuel Roberts: exactly what it was. I think I have some I just…
210 00:25:25.120 ⇒ 00:25:34.869 Uttam Kumaran: Actually, you guys will find it interesting on how I even did, how I even executed the actual thing, so let me see if I can… if I…
211 00:25:35.090 ⇒ 00:25:43.969 Uttam Kumaran: Shit, let me see if I have, the, remotes, oh yeah, so…
212 00:25:44.470 ⇒ 00:25:49.879 Uttam Kumaran: I think you guys will find it interesting how I actually did this. So, I literally said.
213 00:25:50.030 ⇒ 00:26:00.599 Uttam Kumaran: I use the last 30 days command. The last 30 days command is really amazing. What it does is it looks through Reddit, YouTube, X, and, like, the web.
214 00:26:00.790 ⇒ 00:26:04.179 Uttam Kumaran: On, like, what are the last 30-ish days of, like.
215 00:26:04.500 ⇒ 00:26:16.210 Uttam Kumaran: commentary on a subject, and I said, how to use TMUX Tailscale to connect Mac Mini, or your whatever your machine is from anywhere, both mobile and desktop coding. I was just, like, speech-to-text.
216 00:26:16.350 ⇒ 00:26:21.010 Uttam Kumaran: it did a bunch of research, and I said, That looks good.
217 00:26:21.330 ⇒ 00:26:22.969 Uttam Kumaran: Set it up.
218 00:26:23.430 ⇒ 00:26:29.900 Uttam Kumaran: just walked me through how to do it, and I was running to this interview that I had to drive to San Antonio to, so I had, like, 10 minutes.
219 00:26:30.050 ⇒ 00:26:31.859 Uttam Kumaran: And then it worked, and then I just left.
220 00:26:32.340 ⇒ 00:26:45.929 Uttam Kumaran: So, this is another pattern I think you guys should think about using, is if you’re ever, like, hey, I need to do research on a subject that I don’t know, last 30 days is clutch. As you can just look and be like, how… what does the industry think about, like, this… this topic, you know?
221 00:26:48.080 ⇒ 00:26:57.129 Samuel Roberts: Yeah, I played with a few tools for doing something like this when I was traveling, like, a month or two ago, and I… this seems like the best way to do it. There’s, like, a few tools that were trying to let you, like.
222 00:26:57.370 ⇒ 00:27:04.730 Samuel Roberts: connect directly to Cloud Code and give you a GUI in the web and stuff, and I think this is just the cleanest way to do it, from what I was able to find.
223 00:27:04.850 ⇒ 00:27:10.040 Samuel Roberts: Yeah, and that was more than 30 days ago, so I’m sure it’s still, like… Yeah. Cool.
224 00:27:11.480 ⇒ 00:27:12.280 Samuel Roberts: Sweet.
225 00:27:19.280 ⇒ 00:27:27.149 Uttam Kumaran: Okay, cool. So, I mean, yeah, I think give this a shot, like, give the OpenCode CLI a shot, let me know where you get stuck.
226 00:27:27.330 ⇒ 00:27:33.079 Uttam Kumaran: I’ll start sending some updates on how to assign tickets out to OpenCode and things like that.
227 00:27:33.440 ⇒ 00:27:35.540 Uttam Kumaran: And just, like, let me know where you guys get stuck.
228 00:27:35.790 ⇒ 00:27:43.989 Uttam Kumaran: a win for me is that, like, you guys start using this, because… and also, I think my pitch to you is that 5.4, you can use 5.4 on here, so…
229 00:27:45.010 ⇒ 00:27:47.510 Uttam Kumaran: That’s, like, the… why you should…
230 00:27:47.770 ⇒ 00:27:51.080 Uttam Kumaran: That’s my pitch to you on why you should take, like, 10 minutes to figure this out.
231 00:27:51.870 ⇒ 00:27:57.499 Samuel Roberts: What did we get… so there’s 5.4, it looked like there was… what are the other ones we had set up? We had Kimmy set up at one point as well.
232 00:27:57.500 ⇒ 00:28:05.979 Uttam Kumaran: Yeah, I think it’s… the problem is, like, some of the Azure ones don’t have, like, a responses endpoint, so you can use 5.4 and 5.4 mini right now.
233 00:28:05.980 ⇒ 00:28:07.320 Samuel Roberts: Okay Okay.
234 00:28:07.320 ⇒ 00:28:10.590 Uttam Kumaran: And then I’m gonna work… I’ll… probably this weekend, I’ll work on that.
235 00:28:10.830 ⇒ 00:28:12.419 Uttam Kumaran: Adding some more support, yeah.
236 00:28:12.420 ⇒ 00:28:14.049 Samuel Roberts: should be plenty, I feel like, right now, so…
237 00:28:14.050 ⇒ 00:28:15.290 Uttam Kumaran: Yeah, yeah.
238 00:28:15.810 ⇒ 00:28:23.820 Samuel Roberts: Especially if we have credits, I’m not gonna… yeah. The only other thing I’d be worried about is just, like, do I need 5.4 right now? But if there’s credits, I’m not gonna stress about it for a while, I guess.
239 00:28:25.450 ⇒ 00:28:26.170 Samuel Roberts: Cool.
240 00:28:27.110 ⇒ 00:28:32.889 Pranav Narahari: I also feel like, with this process now, that things can be done end-to-end more confidently.
241 00:28:33.190 ⇒ 00:28:37.440 Pranav Narahari: We… as soon as we build out a ticket and we’re happy with it, we should just send it out to…
242 00:28:37.770 ⇒ 00:28:41.149 Pranav Narahari: To, like, in open work, just, like, get it going.
243 00:28:41.700 ⇒ 00:28:42.960 Uttam Kumaran: Yeah. Yes.
244 00:28:44.040 ⇒ 00:29:00.299 Pranav Narahari: what we probably should do, and maybe we’ll learn this as we go, is, like, how do we want those things merged into, GitHub? Like, what was the type of format? We can probably create a playbook specifically for that, so then OpenWork can, like, look at that.
245 00:29:00.420 ⇒ 00:29:14.799 Pranav Narahari: just because we want to, probably, like, on a per milestone basis, figure out which things do we want to merge, which things do we want to QA, and maybe that’s what we do at the beginning of every week. We see, like, okay, what is the gap between what OpenCode did versus where we need to be for our milestone?
246 00:29:14.800 ⇒ 00:29:21.359 Pranav Narahari: And then that’s what we actually end up working on, or refining the ticket to then give it back to OpenCode.
247 00:29:21.460 ⇒ 00:29:25.020 Pranav Narahari: So… Okay, that’s, like, where I’m thinking right now. Yeah.
248 00:29:25.020 ⇒ 00:29:25.530 Samuel Roberts: I like that.
249 00:29:25.700 ⇒ 00:29:26.829 Pranav Narahari: Start using this.
250 00:29:28.340 ⇒ 00:29:43.530 Samuel Roberts: Yeah, I think there was something I was trying to do for Eden and ABC when I was putting together the tickets, Pranav, you know this, where I was like, this is AI, or agent possible, or human-required kind of thing, so, like, if someone needs to be executing on it for, like, any reason, if there’s decisions to be made by a human.
251 00:29:43.600 ⇒ 00:29:59.159 Samuel Roberts: that should be… I wanted to surface that in a label. Otherwise, I wanted to just say, like, maybe an agent could just go off and do it on its own, and see what happens. So I think I’m gonna try to use those labels, at least to make things clear when people get tickets, if they can just kick it off, or if they need to be part of the loop.
252 00:29:59.670 ⇒ 00:30:09.310 Pranav Narahari: Yeah, I guess what I’m saying now, though, is, like, we may not even need those tags anymore. It’s just automatically it goes to the agent.
253 00:30:10.520 ⇒ 00:30:23.179 Samuel Roberts: Sure, sure, oh, you’re right, you’re right. I still think there’s probably a human-required one, just in case there’s, like, decisions to be made, and I want someone to know that, and if that’s linear or open code, it should know that there’s some kind of…
254 00:30:23.620 ⇒ 00:30:29.390 Samuel Roberts: step there. But yeah, you’re right, maybe we don’t even need that. Maybe the AI can just be like, oh, I can’t make this decision anymore, and it’ll surface it.
255 00:30:29.390 ⇒ 00:30:30.200 Pranav Narahari: Yeah, like…
256 00:30:30.200 ⇒ 00:30:30.769 Samuel Roberts: That might be enough.
257 00:30:31.100 ⇒ 00:30:33.640 Pranav Narahari: Fill in the gap before the milestone comes up.
258 00:30:34.500 ⇒ 00:30:41.939 Samuel Roberts: Yeah, yeah, yeah. I’m just thinking ticket-wise, like, if a ticket, like, trying to break apart what’s decisions that need to be made versus the actual execution.
259 00:30:43.130 ⇒ 00:30:44.109 Pranav Narahari: Hmm, okay.
260 00:30:44.110 ⇒ 00:30:46.999 Samuel Roberts: Even if those tickets are decision tickets, kind of thing, you know.
261 00:30:47.000 ⇒ 00:31:01.169 Uttam Kumaran: Yeah, but this is where, like, have judgment. If you think a ticket is kind of finicky, then tell it, like, yo, this ticket’s finicky, like, loop me in where you need me. Yeah. If you think it’s… some of this stuff I’m working on, I’m like, don’t stop, like, this is a pretty straightforward thing, like.
262 00:31:01.170 ⇒ 00:31:01.710 Samuel Roberts: Yeah.
263 00:31:01.710 ⇒ 00:31:03.170 Uttam Kumaran: railway loop.
264 00:31:03.550 ⇒ 00:31:07.129 Uttam Kumaran: loop until… Stuff is resolved.
265 00:31:07.310 ⇒ 00:31:19.829 Uttam Kumaran: And so, I think have some judgment. You can assume that over time, I will… the platform will be set up in a way where you don’t really have to think, but I’m telling you, the more you can have in linear.
266 00:31:20.010 ⇒ 00:31:22.060 Uttam Kumaran: Planned out before, the better.
267 00:31:22.340 ⇒ 00:31:28.490 Uttam Kumaran: and it’s just gonna succeed, like, the odds of success. I’m finding it, because I’m asking…
268 00:31:28.660 ⇒ 00:31:35.050 Uttam Kumaran: I’m stress testing it to do, like, full-length projects with absolutely no context, so…
269 00:31:35.520 ⇒ 00:31:51.630 Uttam Kumaran: I’m telling you, if you even spend half, like, you spend, like, 5 minutes, like, brain dumping, it’s gonna rip it. I’m literally like, yo, implement open panel across these 5 surfaces. Oh, actually, I didn’t even do that. I said, use last 30 days, tell me what the market thinks is the best
270 00:31:51.680 ⇒ 00:32:02.099 Uttam Kumaran: product analytics suite. It said Post Hog or Open Panel. I said, okay, go ahead and build open panel across all of our core internal product surfaces.
271 00:32:02.670 ⇒ 00:32:11.400 Uttam Kumaran: Like, that’s, like, kind of OD. It’s, like, kind of, like, unfair for me to ask it to do that. And then, yeah, and then it’s, like, it’ll… but it still worked, and then at some point.
272 00:32:11.400 ⇒ 00:32:11.870 Samuel Roberts: Yeah.
273 00:32:11.870 ⇒ 00:32:20.290 Uttam Kumaran: it was doing some things, and I’m like, why do we need Click House? But then it’s like, no, Click House is a railway service, like, okay, alright. So, but, like, see, that’s, like, that’s where I’m, like.
274 00:32:20.510 ⇒ 00:32:30.829 Uttam Kumaran: that’s, like, not enough content. Like, I don’t feel like that’s… that was totally fair. But see, for me, I’m gonna push it to figure out, like, okay, what… because I’m assuming that
275 00:32:31.190 ⇒ 00:32:40.729 Uttam Kumaran: not this group, but, like, other people are going to use it with that little context, still needs to succeed, or it needs to push back, right? So I’m trying to figure that out a little bit.
276 00:32:40.730 ⇒ 00:32:41.610 Samuel Roberts: Yeah.
277 00:32:42.670 ⇒ 00:32:53.960 Samuel Roberts: Okay, cool. And then the open panel, that’ll be, like, where we’ll be able to see, like, token usage and things like that, eventually, for open code? Because obviously, like, that’s one nice thing about cursor, is that it does surface a lot of those metrics for us, and I imagine this is.
278 00:32:53.960 ⇒ 00:33:06.269 Uttam Kumaran: Yes, I’m gonna get… I’m gonna get that… I’m gonna get that all through OpenPanel, or I’ll… I’ll use OpenLM or something eventually. And really, my… part of my goal, too, is, like, I can’t see people’s chat experiences with.
279 00:33:07.260 ⇒ 00:33:14.579 Uttam Kumaran: cursor right now, which means I can’t, like, I can’t identify ways to improve the system. I’m mainly gathering that through, like.
280 00:33:14.900 ⇒ 00:33:17.090 Uttam Kumaran: These, like, feedback sessions, you know?
281 00:33:19.110 ⇒ 00:33:19.830 Samuel Roberts: Makes sense.
282 00:33:20.300 ⇒ 00:33:21.010 Samuel Roberts: Sweet.
283 00:33:25.040 ⇒ 00:33:25.460 Uttam Kumaran: Okay.
284 00:33:26.700 ⇒ 00:33:29.019 Uttam Kumaran: Alright guys, thank you, let me know what you think.
285 00:33:29.900 ⇒ 00:33:30.560 Samuel Roberts: Alrighty.
286 00:33:30.850 ⇒ 00:33:31.930 Samuel Roberts: Okay.
287 00:33:31.930 ⇒ 00:33:32.430 Uttam Kumaran: shed.
288 00:33:32.800 ⇒ 00:33:33.560 Uttam Kumaran: Bye.
289 00:33:33.920 ⇒ 00:33:34.400 Mustafa Raja: Thank you.