Meeting Title: Brainforge EP Process Improvement Sync Date: 2026-02-12 Meeting participants: Uttam Kumaran, Brylle Girang, Clarence Stone
WEBVTT
1 00:00:30.350 ⇒ 00:00:36.660 Uttam Kumaran: Yeah, I guess I was saying, just think about it like an intern. You just have to be, like, a little… you have to be just, like.
2 00:00:36.770 ⇒ 00:00:38.180 Uttam Kumaran: very specific.
3 00:00:38.500 ⇒ 00:00:40.730 Uttam Kumaran: And you’ll just find it becomes a muscle, like.
4 00:00:40.870 ⇒ 00:00:45.720 Uttam Kumaran: usually when it fails, it’s because you’re like, damn, I totally didn’t give it enough information, so…
5 00:00:46.310 ⇒ 00:00:47.360 Brylle Girang: I agree.
6 00:00:48.440 ⇒ 00:00:56.170 Uttam Kumaran: Cool. So yeah, maybe continue. I think, kind of primary, like, and, you know, we don’t have to stay on forever, is I just wanted, kind of, for you to share…
7 00:00:56.430 ⇒ 00:00:57.390 Uttam Kumaran: with…
8 00:00:58.430 ⇒ 00:01:06.740 Uttam Kumaran: Clarence and I, like, how it’s going with the structure and cursor, how you’re using the daily meeting logs, and just kind of get your reflection so far, really.
9 00:01:07.350 ⇒ 00:01:23.309 Brylle Girang: Yeah, okay, so when you provided me the onboarding materials, the first thing that I did is I tried to use Cursor to help me understand what my roles and responsibilities are. So instead of me going through the markdowns, instead of me going through the documents.
10 00:01:23.310 ⇒ 00:01:31.259 Brylle Girang: I actually just ask it everything that I need to understand. So what my role is, what are the expectations.
11 00:01:33.000 ⇒ 00:01:37.930 Brylle Girang: what are the daily logs, what do I need to do with that, etc. And that was actually my first step.
12 00:01:38.890 ⇒ 00:01:39.380 Uttam Kumaran: Okay.
13 00:01:39.380 ⇒ 00:01:49.830 Brylle Girang: And then, when I… when I understood everything, I now use Cursor as my actual data notebook, so instead of me
14 00:01:49.830 ⇒ 00:02:04.480 Brylle Girang: going to my devices, Apple Notes, instead of me creating a Google document to access my notepad, I just created a new agent where I specifically asked it to act as my notebook, where I will dump stuff.
15 00:02:04.640 ⇒ 00:02:23.119 Brylle Girang: I will… I will provide ideas, I will provide insights, I just use… I just use the voice transcription, and then just let it formatted in a way that’s going to be understandable for both you and me. And it was here… let me double check, but one…
16 00:02:42.900 ⇒ 00:02:44.770 Brylle Girang: Actually, yeah.
17 00:02:50.090 ⇒ 00:02:51.050 Brylle Girang: Okay.
18 00:02:52.330 ⇒ 00:02:54.040 Clarence Stone: Right here, unnamed seconds.
19 00:02:55.570 ⇒ 00:03:00.079 Brylle Girang: Yeah, but I don’t think that I named it specifically as my notebook.
20 00:03:01.980 ⇒ 00:03:03.869 Brylle Girang: Okay, just gonna ask her, sir.
21 00:03:06.400 ⇒ 00:03:07.560 Brylle Girang: in me.
22 00:03:12.460 ⇒ 00:03:18.979 Clarence Stone: I’ve actually never looked at the session logs. Did they preserve everything, even when they compressed context?
23 00:03:21.140 ⇒ 00:03:25.850 Uttam Kumaran: I don’t think it’s gonna put the context in there. I think it’s just gonna put the question-answer pair.
24 00:03:27.710 ⇒ 00:03:33.499 Clarence Stone: No, no, I mean, when your chat window compresses context, you know that circle on the bottom right?
25 00:03:33.500 ⇒ 00:03:34.100 Uttam Kumaran: Yeah.
26 00:03:35.330 ⇒ 00:03:39.060 Clarence Stone: Does that MD file also compress, or is it just a log?
27 00:03:40.740 ⇒ 00:03:45.630 Uttam Kumaran: No, dude, I think it’s just a log. Like, you can’t… you’re not gonna… I don’t think you can recreate, like…
28 00:03:46.120 ⇒ 00:03:46.680 Uttam Kumaran: what it…
29 00:03:46.680 ⇒ 00:03:47.060 Clarence Stone: Oh.
30 00:03:47.060 ⇒ 00:03:48.680 Uttam Kumaran: I don’t think you could recreate end-to-end.
31 00:03:51.650 ⇒ 00:03:59.470 Uttam Kumaran: Like, I don’t think you can take a session and be like, I want to walk through this exact, like, the context it pulled and everything. I think it’s just the QA pairs.
32 00:04:00.910 ⇒ 00:04:06.569 Clarence Stone: Oh, I mean, that’s good enough. Yeah. I was just worried that it wouldn’t save it if it compressed contacts.
33 00:04:06.570 ⇒ 00:04:10.450 Uttam Kumaran: Oh… no, I think… I think it’s separating it out.
34 00:04:11.360 ⇒ 00:04:16.210 Uttam Kumaran: Like, your session, your QA pairs versus… and the outputs versus, like.
35 00:04:17.490 ⇒ 00:04:20.339 Uttam Kumaran: The max context is all factored into the 100%.
36 00:04:20.540 ⇒ 00:04:25.940 Clarence Stone: Yeah, it’s a pain in the ass to write code to figure out what the thinking is. It was really painful.
37 00:04:26.190 ⇒ 00:04:30.920 Clarence Stone: But it’s… it’s not worth it. That… you don’t need to.
38 00:04:30.920 ⇒ 00:04:33.469 Uttam Kumaran: Yeah, dude, you just gotta drag the…
39 00:04:33.470 ⇒ 00:04:34.850 Clarence Stone: Fuck.
40 00:04:35.200 ⇒ 00:04:36.150 Uttam Kumaran: You got anything? Yeah.
41 00:04:36.150 ⇒ 00:04:40.169 Clarence Stone: Wait, so tell it? Like, what else did you forget?
42 00:04:40.880 ⇒ 00:04:52.290 Clarence Stone: No, no, like, I don’t stop there. I’m like, okay, what else, like, what else did you forget? Like, you owe me all my sessions, right? We had an agreement. What’s going on?
43 00:04:53.330 ⇒ 00:05:02.209 Brylle Girang: And we asked that what, did you miss anything from the… Into Millions.
44 00:05:02.680 ⇒ 00:05:14.040 Uttam Kumaran: Okay, but let’s keep going, because I have to… I have to drive. I’ve been on, like, I’ve been on, like, 6 hours of meetings, but I just… Bea, I just wanted you to give a sort of update on, like, yeah, how… how it’s been going since…
45 00:05:14.220 ⇒ 00:05:15.850 Uttam Kumaran: Onboarding, and then, like.
46 00:05:16.380 ⇒ 00:05:26.069 Uttam Kumaran: just kind of give us some feedback on, like, yeah, how this cursor setup is going, and you had two main kind of tasks, the element thing and this today, so if you could just, like, reflect on that.
47 00:05:29.160 ⇒ 00:05:42.569 Brylle Girang: Gotcha. So, when I’m thinking about the adoption, like, what’s going to be the friction between our people and then Cursor, I look back into how you taught me to set this up.
48 00:05:42.880 ⇒ 00:05:55.870 Brylle Girang: And I was thinking, is that going to be the same thing for all other employees? If you needed… if you needed to teach me hands-on, do we need to do
49 00:05:55.970 ⇒ 00:06:00.670 Brylle Girang: that same thing to all other people. So, I was checking…
50 00:06:01.440 ⇒ 00:06:19.399 Brylle Girang: any documentations, any guides, and I have seen that we do have those in the GitHub repo, but I do think that that’s going to be one of the main friction points when it comes to adopting cursor. But overall, after that, it was fairly easy.
51 00:06:19.680 ⇒ 00:06:23.669 Brylle Girang: Well, I have experience with VS Code, I have experience with
52 00:06:24.150 ⇒ 00:06:34.090 Brylle Girang: well, using ChatGPT almost every day before, before my role here, so I’m not sure if that gave me an edge into actually adopting cursor.
53 00:06:34.200 ⇒ 00:06:37.230 Brylle Girang: but when it comes to the element…
54 00:06:37.230 ⇒ 00:06:41.669 Uttam Kumaran: I think take out your enthusiasm, because that’s what not everybody’s gonna have.
55 00:06:41.880 ⇒ 00:06:42.200 Brylle Girang: Yeah.
56 00:06:42.200 ⇒ 00:06:45.969 Uttam Kumaran: is, like, your drive and your curiosity. So, a discount for that.
57 00:06:46.270 ⇒ 00:06:49.490 Uttam Kumaran: It’s sort of like the people that we’re thinking about.
58 00:06:49.930 ⇒ 00:06:50.810 Uttam Kumaran: You know?
59 00:06:52.630 ⇒ 00:06:54.120 Brylle Girang: Yeah.
60 00:06:54.120 ⇒ 00:06:56.900 Uttam Kumaran: Cause some people are gonna be like, oh, I just, like, it didn’t work.
61 00:06:57.800 ⇒ 00:06:58.310 Uttam Kumaran: Like.
62 00:06:58.310 ⇒ 00:07:09.310 Brylle Girang: Yeah, I do agree with that, because, you know, if you give the people maybe a harder path to follow the first time around, they will automatically default to the
63 00:07:09.550 ⇒ 00:07:15.910 Brylle Girang: Same thing that they have been doing for the past few years. I think that’s also one of the main friction points.
64 00:07:16.800 ⇒ 00:07:17.380 Uttam Kumaran: Okay.
65 00:07:19.340 ⇒ 00:07:30.350 Brylle Girang: With setting the cursor, with the difficulty… maybe the difficulty of setting cursor one time from the start, that’s going to, like, push
66 00:07:30.610 ⇒ 00:07:37.660 Brylle Girang: all other people into not using cursor. Because if it’s not easy the first time around.
67 00:07:38.240 ⇒ 00:07:45.910 Brylle Girang: they might think that it’s not going to be easy when they use it permanently. So, that’s just… that’s just my idea. And then…
68 00:07:47.270 ⇒ 00:07:55.529 Brylle Girang: At the end of the day, I do think that there are so many options right now that people default to the easier
69 00:07:55.790 ⇒ 00:07:56.490 Brylle Girang: F?
70 00:07:56.630 ⇒ 00:08:04.399 Brylle Girang: for example, if I want to get transcripts, if I want to get the recordings, we have Cursor, but at the same time, we have the platform.
71 00:08:04.400 ⇒ 00:08:05.170 Uttam Kumaran: Yeah.
72 00:08:05.170 ⇒ 00:08:14.950 Brylle Girang: scan through the videos, they can read through the transcripts by themselves, they can copy the transcripts and paste it into whatever AI tool they’re using, ChatGPT, etc.
73 00:08:15.080 ⇒ 00:08:16.120 Brylle Girang: So…
74 00:08:17.260 ⇒ 00:08:29.450 Brylle Girang: may be the… the harshest thing that we can do in the future, just in case this doesn’t work, but the harshest thing that we may need to do in the future is just reduce the options.
75 00:08:29.580 ⇒ 00:08:30.680 Brylle Girang: And then…
76 00:08:30.940 ⇒ 00:08:31.560 Uttam Kumaran: Oh, okay.
77 00:08:31.560 ⇒ 00:08:32.520 Brylle Girang: everyone.
78 00:08:32.679 ⇒ 00:08:37.289 Brylle Girang: to funnel their efforts into Cursor. Like, make Cursor the go-to.
79 00:08:37.690 ⇒ 00:08:39.770 Brylle Girang: Shop for everyone else.
80 00:08:42.530 ⇒ 00:08:49.970 Brylle Girang: I think the hardest task for me was the element weekly, weekly update, just because I don’t have
81 00:08:50.270 ⇒ 00:08:55.899 Brylle Girang: enough context to build the ideal tech that I want to build.
82 00:08:55.990 ⇒ 00:09:14.739 Brylle Girang: We have the transcripts, but again, I think I mentioned to you, that I’m not sure if I’m not getting all the resources, but I only had transcripts prior to February, and we needed to create an update that’s going to be for the last two weeks. That was the main challenge, and at the same time.
83 00:09:15.080 ⇒ 00:09:17.480 Brylle Girang: I don’t have a way to just…
84 00:09:18.090 ⇒ 00:09:23.530 Brylle Girang: Understand in one go what we’re doing for the client, where are we.
85 00:09:23.890 ⇒ 00:09:27.420 Brylle Girang: For this client, what are we missing?
86 00:09:27.420 ⇒ 00:09:33.720 Uttam Kumaran: Can you explain… I think… can you explain to Clarence, like, what the task was today? And then, I think that’s what I really wanted him to hear, because…
87 00:09:33.870 ⇒ 00:09:39.290 Uttam Kumaran: Clarence, one thing that we tried to ask people, especially the EPs, is, like, to create a client hub.
88 00:09:39.580 ⇒ 00:09:47.540 Uttam Kumaran: And really, like, that’s what prevented B from, like, taking this to the, like, 100% today, is, like, there wasn’t a clear client hub, and, like.
89 00:09:47.740 ⇒ 00:09:53.479 Uttam Kumaran: For this client, and so he’s purely going off of, like, whatever transcripts and, like, code was in
90 00:09:54.300 ⇒ 00:09:55.140 Uttam Kumaran: the repo?
91 00:09:55.140 ⇒ 00:10:01.520 Brylle Girang: Yeah, yeah, totally agree. So, we have several client resources, but they’re scattered.
92 00:10:01.790 ⇒ 00:10:19.310 Brylle Girang: So we don’t have, like, a unified place to look at the client and then understand everything that’s going on. I know that we have Instagant to check the overall project timeline, we have Linear to check the individual tickets, and then we have Notion for some client info.
93 00:10:19.420 ⇒ 00:10:27.279 Brylle Girang: And then those are not connected yet to Courser. So I needed to look at those individually, just to understand where we are at.
94 00:10:27.460 ⇒ 00:10:31.659 Clarence Stone: Okay, yeah, so I think next step is you should get your MCPs connected.
95 00:10:32.830 ⇒ 00:10:36.329 Clarence Stone: We can get you… like, Cursor can look at those things through MCP.
96 00:10:37.510 ⇒ 00:10:44.190 Uttam Kumaran: Not… I don’t think the Notion… yeah, I think… but the Notion MC… like, I think that’s the thing we’re just gonna have to test, like, whether it’s all working.
97 00:10:44.350 ⇒ 00:10:49.089 Uttam Kumaran: Because the Notion MCP craps out, like, if it doesn’t know the ontology.
98 00:10:49.090 ⇒ 00:10:51.079 Clarence Stone: Fragile. Very fragile.
99 00:10:51.230 ⇒ 00:10:51.810 Uttam Kumaran: Yeah.
100 00:10:55.310 ⇒ 00:10:57.919 Clarence Stone: It’s gone to great lengths to try and fix it.
101 00:11:00.250 ⇒ 00:11:02.400 Uttam Kumaran: So, I think, like, I don’t… yeah, go ahead, go ahead.
102 00:11:02.840 ⇒ 00:11:22.690 Brylle Girang: Sorry. Oh, when it comes to actually creating the deck, I think that the recent PowerPoint MCPs that you shared, Etan, that would be a lot helpful. Okay. So if you could just, you know, if we can get a way to connect Instagan, to connect Linear, to connect Notion into Cursor, and then just
103 00:11:22.860 ⇒ 00:11:24.370 Brylle Girang: Create a plan.
104 00:11:24.750 ⇒ 00:11:29.420 Brylle Girang: To… to, like, update the deck instantly, then that would be amazing.
105 00:11:29.960 ⇒ 00:11:34.439 Brylle Girang: So that reduces maybe 30 minutes, 60 minutes of our time.
106 00:11:34.440 ⇒ 00:11:35.389 Uttam Kumaran: per week. Okay.
107 00:11:35.730 ⇒ 00:11:40.520 Brylle Girang: To allow us to… to dedicate that to more strategic things, right?
108 00:11:40.740 ⇒ 00:11:41.180 Uttam Kumaran: Yeah.
109 00:11:41.750 ⇒ 00:11:43.980 Brylle Girang: The case study was…
110 00:11:44.720 ⇒ 00:11:56.609 Brylle Girang: easier than the element task, just because I have everything that I need. I have the template, I have the prompt from the marketing team, I have the why, and that helps
111 00:11:57.270 ⇒ 00:12:01.509 Brylle Girang: That really helps when it comes to creating the plan for Cursor.
112 00:12:02.780 ⇒ 00:12:05.420 Brylle Girang: So yeah, what else?
113 00:12:05.680 ⇒ 00:12:15.380 Clarence Stone: So I’ll also note that Slack seems to be very rich with great context, and it rarely makes it to, like, the, like, more permanent platforms.
114 00:12:15.380 ⇒ 00:12:26.330 Clarence Stone: And that’s been, like, feedback that I get from EPs and service leaders saying, like, hey, we’re having a lot of conversations, I don’t see the tickets being updated.
115 00:12:26.960 ⇒ 00:12:29.500 Clarence Stone: Right? And if you’re a service leader for, like.
116 00:12:29.500 ⇒ 00:12:30.020 Uttam Kumaran: Yes.
117 00:12:30.020 ⇒ 00:12:39.969 Clarence Stone: 7 projects, it just gets really confusing. You don’t remember which chat it was in, you know, the… you know, you have to scroll up really far, and the tickets aren’t updated.
118 00:12:40.610 ⇒ 00:12:45.290 Uttam Kumaran: Well, a lot of people are not, after a meeting, going into Kirscher and saying.
119 00:12:45.460 ⇒ 00:12:48.990 Uttam Kumaran: look at my last meeting and help me create the tickets for it. They’re not doing that.
120 00:12:48.990 ⇒ 00:12:49.560 Brylle Girang: Yeah.
121 00:12:50.150 ⇒ 00:12:50.770 Uttam Kumaran: Right?
122 00:12:51.160 ⇒ 00:13:02.570 Uttam Kumaran: I’m not doing that because I’m in 6 hours of meetings a day. They’re not doing it. There’s no… I don’t know, there’s no reason. They’re not doing it. I don’t know what the reason could be. I don’t… I don’t… I don’t know.
123 00:13:02.880 ⇒ 00:13:11.539 Uttam Kumaran: So, that’s something B, is, like, it’s… that’s what you… I want you to kind of see in this next two weeks, is that, like, all of the things an EP has to do.
124 00:13:12.380 ⇒ 00:13:15.799 Uttam Kumaran: What is the friction in between them doing it on Kirscher today?
125 00:13:16.230 ⇒ 00:13:19.659 Brylle Girang: Let’s remove that friction as much as we can.
126 00:13:19.730 ⇒ 00:13:24.829 Uttam Kumaran: So that you can then call each of them and say, guys, There’s no more friction.
127 00:13:25.780 ⇒ 00:13:29.219 Uttam Kumaran: There’s no reason for them… for linear tickets to be out of date.
128 00:13:29.530 ⇒ 00:13:34.460 Uttam Kumaran: If you raise your hand, if you don’t know how to use Cursor to do that, I will teach you how to do that.
129 00:13:35.100 ⇒ 00:13:40.210 Uttam Kumaran: Otherwise, if you do know, then there’s no longer an excuse, because the sort of hammer’s gonna start coming down, like…
130 00:13:41.350 ⇒ 00:13:46.009 Uttam Kumaran: You know? And to tell… to give you a sense, we’re gonna start to tie bonuses
131 00:13:46.390 ⇒ 00:13:50.790 Uttam Kumaran: To hitting things like… Take it freshness.
132 00:13:50.900 ⇒ 00:13:52.080 Uttam Kumaran: And stuff like that.
133 00:13:52.250 ⇒ 00:13:56.999 Uttam Kumaran: So, Clarence, like, one of the things we worked on is that for EPs, at any moment.
134 00:13:57.890 ⇒ 00:14:02.299 Uttam Kumaran: Clarence and I can be like, okay, for your client, we want to see all the resources.
135 00:14:03.580 ⇒ 00:14:07.259 Uttam Kumaran: And if you can’t produce that, I don’t know, Clint, what did we say? If you can’t produce that.
136 00:14:07.400 ⇒ 00:14:12.890 Uttam Kumaran: Or, like, what do we say if we can’t just go to one place to, like, and be like, cool, this is good? You failed.
137 00:14:12.890 ⇒ 00:14:30.589 Clarence Stone: Yeah, so, I have a list of questions that I ask, and it’s like business-type questions. It’s like, hey, give me an update, about your, project. What’d you guys do last week? What did you talk about with the client, right? And if they can’t show me where I can find that information, then we have problems.
138 00:14:32.300 ⇒ 00:14:49.299 Clarence Stone: Right, that’s how I test whether or not, like, they’re actually storing it. I personally don’t care where they store it. You saw it just now, it’s not that hard for us to get all the connections, but it’s not okay if you’re not logging it somewhere, because Slack is the most ephemeral of all the feeds that we have for communication.
139 00:14:49.600 ⇒ 00:14:52.330 Clarence Stone: Right, it’s the most noisy, it’s gonna be the most expensive.
140 00:14:52.590 ⇒ 00:14:54.050 Clarence Stone: It can’t stay there.
141 00:14:54.740 ⇒ 00:14:55.300 Uttam Kumaran: Yeah.
142 00:14:55.460 ⇒ 00:15:01.539 Uttam Kumaran: So, B, what you’re seeing on my… on the clients that I’m involved in, you’re gonna see the most disorganization.
143 00:15:02.070 ⇒ 00:15:12.230 Uttam Kumaran: Because I’m involved. So, like, I’m… because I’m really good at my job, I don’t need that much organization to run these clients. Because I can…
144 00:15:12.370 ⇒ 00:15:22.309 Uttam Kumaran: this is, like, well, all I do. So, like, I can go in and dance, and I… but for our other cus… for other clients, and for mine, too, that is not the standard.
145 00:15:22.470 ⇒ 00:15:26.969 Uttam Kumaran: Because not everybody… we can’t rely on everybody being able to…
146 00:15:27.490 ⇒ 00:15:30.220 Uttam Kumaran: Dance, and sort of figure it out as they go.
147 00:15:31.010 ⇒ 00:15:41.689 Uttam Kumaran: like, I just do this for a living, and I’ve done this now for, like, a lot, so I’m able to kind of patch things together and get it going. Not everybody’s able to do that. So, to Clarence’s point, at any moment.
148 00:15:41.820 ⇒ 00:15:46.500 Uttam Kumaran: For us to be able to, like, here are 10 questions, we need to answer this, you have 30 minutes.
149 00:15:46.970 ⇒ 00:15:48.940 Uttam Kumaran: For Element, it would have been a fail.
150 00:15:49.150 ⇒ 00:15:49.950 Uttam Kumaran: Right?
151 00:15:50.090 ⇒ 00:15:50.490 Brylle Girang: Yep.
152 00:15:50.490 ⇒ 00:15:54.179 Uttam Kumaran: Which is valid, because we don’t have an EP on this clock. That’s me.
153 00:15:54.450 ⇒ 00:15:57.879 Uttam Kumaran: So, but for the other clients, right? So, if we were to go to Eden.
154 00:15:58.010 ⇒ 00:16:07.300 Uttam Kumaran: If we were to go to Magic Spoon, if we were to go to ABC, if you were to call that EP today, I don’t… I don’t think…
155 00:16:08.590 ⇒ 00:16:11.440 Uttam Kumaran: maybe one client, I think they would pass.
156 00:16:12.300 ⇒ 00:16:17.799 Uttam Kumaran: So… One thing that I think, Clarence, for us to think about, especially for B, is, like.
157 00:16:18.170 ⇒ 00:16:22.849 Uttam Kumaran: what is his OKR? Like, is it… 90% pass.
158 00:16:23.410 ⇒ 00:16:31.380 Uttam Kumaran: Right? Or, like, I’m trying to… I’m even trying to think about, like, for the delivery team, what is our… for delivery leadership, what is… what is my OKR?
159 00:16:31.710 ⇒ 00:16:37.630 Uttam Kumaran: that, like, 90% of EPs pass the… clearance check.
160 00:16:38.470 ⇒ 00:16:40.909 Uttam Kumaran: like, you know, I’m trying to give us goals.
161 00:16:41.160 ⇒ 00:16:41.690 Clarence Stone: Yeah.
162 00:16:41.690 ⇒ 00:16:43.489 Uttam Kumaran: so that I can start to…
163 00:16:43.690 ⇒ 00:16:48.109 Uttam Kumaran: hold us accountable for, like, making sure this happens, right? And so that’s sort of… this is, like, what…
164 00:16:48.390 ⇒ 00:16:52.829 Uttam Kumaran: I just wanted to focus on EP, but this is what I want to go do across all the different
165 00:16:53.260 ⇒ 00:16:57.699 Uttam Kumaran: parts of delivery. The services, and each of the roles.
166 00:16:59.980 ⇒ 00:17:06.699 Clarence Stone: Yeah, I don’t… I wouldn’t be happy with 90. It’s like…
167 00:17:06.839 ⇒ 00:17:18.229 Clarence Stone: So, by the way, that… I’m gonna pull up the document real quick. The document’s formatted in a way where I say, like, these 10 questions are easy. You have to get them.
168 00:17:18.500 ⇒ 00:17:24.530 Clarence Stone: I think you need to be able to get the easies, because all of those questions are, where are you storing this?
169 00:17:24.670 ⇒ 00:17:26.539 Clarence Stone: What do you do when this happens?
170 00:17:26.880 ⇒ 00:17:27.770 Clarence Stone: Right.
171 00:17:29.300 ⇒ 00:17:42.180 Clarence Stone: for the more, like, complicated ones, I don’t care if they get them right or wrong, it’s like, are you thinking about, you know, your function as an EP? So, yeah, I think for things that are, like.
172 00:17:43.240 ⇒ 00:17:46.929 Clarence Stone: almost required for your business OS to enrich contacts, UTAM.
173 00:17:47.450 ⇒ 00:17:51.240 Clarence Stone: We’ve got to… to really shoot for that 100, or else, like…
174 00:17:51.240 ⇒ 00:17:55.340 Uttam Kumaran: Yeah, yeah, so that’s fine. And so this is where, for us, how do we do that, Bea?
175 00:17:55.700 ⇒ 00:17:58.700 Uttam Kumaran: There’s… there’s 3 things we always talk to clients about this, too.
176 00:17:59.120 ⇒ 00:18:01.769 Uttam Kumaran: There’s people, there’s process, and there’s tools.
177 00:18:02.090 ⇒ 00:18:06.670 Uttam Kumaran: Right? And so we are going to systematically cross off all the excuses.
178 00:18:07.170 ⇒ 00:18:11.589 Uttam Kumaran: And at some point, if people is a problem, there will be things we could do about that.
179 00:18:11.790 ⇒ 00:18:16.559 Uttam Kumaran: But I am not yet confident that we have nailed the process and the tools.
180 00:18:16.860 ⇒ 00:18:19.189 Uttam Kumaran: So our job is to cross those off.
181 00:18:19.470 ⇒ 00:18:22.899 Uttam Kumaran: then I can go be like, yo, there’s no excuses, right?
182 00:18:23.270 ⇒ 00:18:24.180 Uttam Kumaran: So…
183 00:18:24.480 ⇒ 00:18:29.709 Uttam Kumaran: that’s, like, sort of the… I want to tie a bow on this, is, like, that’s what we’re gonna do for the EPs.
184 00:18:30.100 ⇒ 00:18:34.629 Uttam Kumaran: And I think, really, my question to you is…
185 00:18:37.110 ⇒ 00:18:41.680 Uttam Kumaran: Do we think that we could do, like, an EP month?
186 00:18:42.340 ⇒ 00:18:45.440 Uttam Kumaran: and then we can move on to CSOs next month? Like, what do you think?
187 00:18:47.960 ⇒ 00:18:50.099 Uttam Kumaran: I know, I don’t know, Clarence, like, what do you think? Like.
188 00:18:50.320 ⇒ 00:18:52.959 Uttam Kumaran: I’m almost like, maybe we do 2 weeks.
189 00:18:52.960 ⇒ 00:18:54.359 Clarence Stone: So…
190 00:18:54.730 ⇒ 00:19:02.320 Clarence Stone: I’ll just throw this out there, like, it’s a little different, like, I would want to make sure that we have all the pipelines.
191 00:19:02.460 ⇒ 00:19:06.249 Clarence Stone: And we’re monitoring the pipelines, and then…
192 00:19:06.510 ⇒ 00:19:20.839 Clarence Stone: start with EPs, I agree. Because I think you can do EPs very well in a week if you… if you, you know, set up a standard for how you’re going to fetch that data, right? And then you, like, effectively, you’re testing what you’ve set up.
193 00:19:21.270 ⇒ 00:19:22.130 Uttam Kumaran: on everybody.
194 00:19:22.130 ⇒ 00:19:22.800 Clarence Stone: Do you think?
195 00:19:22.910 ⇒ 00:19:39.170 Clarence Stone: Yeah. And, like, this is just a fun, like, test that I’ve been doing, like, dude, these MCP pipelines, they are incredibly fragile, and I had to go through all of this effort to build this dashboard that would actually log
196 00:19:39.200 ⇒ 00:19:48.549 Clarence Stone: all the issues that happen with MCP connectors. These are all the active subagent-like swarms that are running right now, and
197 00:19:48.800 ⇒ 00:19:54.700 Clarence Stone: When you go down, I do a health monitor check every hour on uptime.
198 00:19:54.810 ⇒ 00:19:56.649 Clarence Stone: For all my… my tools.
199 00:19:58.100 ⇒ 00:20:06.900 Clarence Stone: And when it doesn’t work, I automatically have a loop saying, like, report the issue, and then send a message to Windsorf to fix it.
200 00:20:07.770 ⇒ 00:20:12.939 Clarence Stone: So it does the investigation, sends and prompts Windsorf automatically, and gets a pipeline back up.
201 00:20:12.940 ⇒ 00:20:14.140 Uttam Kumaran: So it stays up, yeah.
202 00:20:14.140 ⇒ 00:20:21.210 Clarence Stone: Yeah, and then it sends me a personal email of all the crash logs that have happened. Look at all this. Like, I have tons of config errors.
203 00:20:23.280 ⇒ 00:20:25.580 Clarence Stone: And here’s my gateway check.
204 00:20:26.470 ⇒ 00:20:38.389 Clarence Stone: Right? I spent zero time making this look pretty… like, it was a single-shot prompt with a list of all the stuff I needed logged, right? And I used OpenTelemetry to do it.
205 00:20:38.450 ⇒ 00:20:50.459 Clarence Stone: So, personally, I think your job’s gonna be a lot easier if you know that when you ask for something from Notion, you ask for something from Millenio, you ask something from the Slack, that you’re gonna get all that information back.
206 00:20:50.790 ⇒ 00:20:51.700 Clarence Stone: Right.
207 00:20:51.830 ⇒ 00:20:56.669 Clarence Stone: Because later, like, if you’re gonna start building, you’re like, you’re not knowing if you’re getting all of it or not.
208 00:21:01.150 ⇒ 00:21:02.140 Uttam Kumaran: Hmm…
209 00:21:04.920 ⇒ 00:21:17.039 Clarence Stone: That’s… I think that’s the first concern, right? Like, because, like, we can say you guys have to have 100% data entry, but if we can’t even fetch it all, will it help you build it?
210 00:21:20.460 ⇒ 00:21:23.340 Uttam Kumaran: So maybe we think about this in a few phases.
211 00:21:24.010 ⇒ 00:21:26.539 Uttam Kumaran: Because I would like to take this one step further.
212 00:21:27.610 ⇒ 00:21:33.089 Uttam Kumaran: I think one is we figure out, B, how do we get everybody on the EP side?
213 00:21:33.200 ⇒ 00:21:36.430 Uttam Kumaran: to just… hit 100%.
214 00:21:37.800 ⇒ 00:21:41.210 Uttam Kumaran: I’m almost wanna go one step further, and like…
215 00:21:42.570 ⇒ 00:21:45.320 Uttam Kumaran: How can we just build, like, an EP agent?
216 00:21:46.790 ⇒ 00:21:50.569 Uttam Kumaran: Right. How do we just build something that EPs a project?
217 00:21:51.260 ⇒ 00:21:53.349 Uttam Kumaran: Without needing a human involved.
218 00:21:53.800 ⇒ 00:21:55.330 Uttam Kumaran: And so I would like you to hold
219 00:21:55.960 ⇒ 00:21:57.940 Uttam Kumaran: I’d just like to pause there, like…
220 00:21:58.450 ⇒ 00:22:05.180 Uttam Kumaran: ask me some questions about that, just so I’m… so, like, B, you’re, like, you feel like you grasp what I just said.
221 00:22:06.650 ⇒ 00:22:12.520 Brylle Girang: Yeah, I do agree with us going through the EP side first.
222 00:22:12.800 ⇒ 00:22:15.609 Brylle Girang: In my current state.
223 00:22:15.760 ⇒ 00:22:21.250 Brylle Girang: well, I would need time to know more about what they’re doing, but when it comes to the…
224 00:22:21.520 ⇒ 00:22:39.600 Brylle Girang: the easier tasks, I would say, updating the decks, knowing the tickets for our clients, knowing what their current statuses are, those are really things that we can leverage cursor on, and maybe those are some questions that we can use as part of the 100%
225 00:22:39.670 ⇒ 00:22:43.189 Brylle Girang: Okay. I agree. EP first.
226 00:22:43.340 ⇒ 00:22:49.189 Brylle Girang: We tried to build them the systems that would make it easier for them to know
227 00:22:49.350 ⇒ 00:22:50.899 Brylle Girang: Where the clients are at.
228 00:22:51.070 ⇒ 00:22:52.819 Brylle Girang: At any moment in time.
229 00:22:54.320 ⇒ 00:22:54.830 Uttam Kumaran: And then tell me.
230 00:22:54.830 ⇒ 00:22:55.210 Brylle Girang: I think.
231 00:22:55.320 ⇒ 00:22:57.740 Uttam Kumaran: You kind of grasping the next phase?
232 00:22:58.310 ⇒ 00:23:02.800 Uttam Kumaran: It’s like, once we’ve built the systems to get all that information.
233 00:23:03.320 ⇒ 00:23:07.290 Uttam Kumaran: as a CSO, I should just be like, Hey.
234 00:23:08.010 ⇒ 00:23:11.409 Uttam Kumaran: Cursor, just do the… do the job of the EP today.
235 00:23:13.690 ⇒ 00:23:14.610 Uttam Kumaran: You kind of see what I’m saying?
236 00:23:14.610 ⇒ 00:23:16.979 Brylle Girang: Yeah, yeah, yeah.
237 00:23:16.980 ⇒ 00:23:19.629 Uttam Kumaran: Now that we’ve… now that… because, ultimately.
238 00:23:19.760 ⇒ 00:23:23.149 Uttam Kumaran: There’s a… the limiting factor here is there’s a human being.
239 00:23:23.330 ⇒ 00:23:28.459 Uttam Kumaran: That has to be, like, okay, I need to update linear, right? But what is informing them to do that?
240 00:23:29.300 ⇒ 00:23:40.849 Uttam Kumaran: It’s a post-meeting, so every time post-meeting, you have to update linear. That sounds like a workflow, right? And so what I want you to think a little bit further about, and you’re just gonna have to sleep on this.
241 00:23:41.160 ⇒ 00:23:47.590 Uttam Kumaran: But what I want you to think about is, if we’ve built the system in a way where all of these things
242 00:23:47.700 ⇒ 00:23:49.489 Uttam Kumaran: can be connected.
243 00:23:49.660 ⇒ 00:23:54.089 Uttam Kumaran: you can talk to Gint, you can create slides, you can go to Linear, you can get Slack.
244 00:23:54.500 ⇒ 00:23:58.919 Uttam Kumaran: Naturally, we should be able to build an agent that just does that.
245 00:23:59.610 ⇒ 00:24:00.400 Uttam Kumaran: Right?
246 00:24:01.100 ⇒ 00:24:05.629 Uttam Kumaran: And so what I want to challenge you is how far we are from that reality.
247 00:24:05.840 ⇒ 00:24:06.620 Uttam Kumaran: You know?
248 00:24:09.210 ⇒ 00:24:28.909 Clarence Stone: Oh, let’s, let’s, okay, sorry, you can go, sorry. I was gonna say, like, can I show you what that agent stuff looks like? I guess, because, like, there’s the connection, right? Like, you’re seeing it from one direction, where you have all of this tooling and data and enrichment and stuff that’s happening, but we want to standardize it into an agent.
249 00:24:28.960 ⇒ 00:24:31.260 Clarence Stone: The fun part is, like, the last…
250 00:24:31.780 ⇒ 00:24:36.879 Clarence Stone: 3 weeks now, gosh. All I was doing was building agents that go the other way.
251 00:24:38.380 ⇒ 00:24:39.290 Uttam Kumaran: Yeah, so maybe…
252 00:24:39.290 ⇒ 00:24:40.480 Clarence Stone: started with Agent.
253 00:24:40.480 ⇒ 00:24:53.349 Uttam Kumaran: Maybe, Clarence, you just show… just show him, like, what it’s like. I think it’s gonna take you a sec to just, like, get to this point, but I don’t want to, like, inundate you, but I wanna show you that that’s the natural progression here.
254 00:24:53.350 ⇒ 00:25:03.999 Clarence Stone: this is… this is probably pretty new stuff. Like, so this is Lola. Lola’s my innovation consulting person. This is all the, like, research.
255 00:25:05.070 ⇒ 00:25:16.000 Clarence Stone: like, analysis, consulting work that I do that I offload to her, right? So Lola has an identity, she has a role, she…
256 00:25:16.000 ⇒ 00:25:25.379 Clarence Stone: has a clear identification of who she reports to. She has a directive from an orchestration agent that actually manages all the routing for her.
257 00:25:26.100 ⇒ 00:25:41.849 Clarence Stone: There’s a personality. I actually let the AI generate this, I don’t really care about that. The interactive model is that I can talk one-to-one to Lola, or it goes through my orchestration agent, right? And there’s some very clear directives. Like, you always start your session with Kim and Cape 2.5.
258 00:25:41.980 ⇒ 00:25:47.040 Clarence Stone: You don’t have to go through Vic, you talk to me when you’re done with your work, right?
259 00:25:47.310 ⇒ 00:25:50.519 Clarence Stone: So, like, you’re… you start with a real human.
260 00:25:51.070 ⇒ 00:25:53.999 Clarence Stone: I started by crafting the agent.
261 00:25:54.160 ⇒ 00:26:05.630 Clarence Stone: What we’re trying to do is connect all of that, right? So, how would an EP interact with people? How would an EP, you know, report up, right? Who would an EP report up to?
262 00:26:06.120 ⇒ 00:26:13.769 Clarence Stone: Right? Who, who, who do they orchestrate things with? Well, that’s the SLs, right, and the CSOs.
263 00:26:13.980 ⇒ 00:26:27.870 Clarence Stone: Right? Like, even though this is an agent, we’re putting in the same parameters that we would for a human. Right? So, like, when, you know, starting, ending, whatever, command modes, right? Like, I can…
264 00:26:28.170 ⇒ 00:26:35.559 Clarence Stone: trigger workflows. So, if I want Lola to do research, I’ll do TAC research and just type in whatever I want flow of thought.
265 00:26:35.840 ⇒ 00:26:43.479 Clarence Stone: Right, and there’s a workflow for research where after she’s done, she… she will post it up to the research folder, and then email me.
266 00:26:44.620 ⇒ 00:27:01.259 Clarence Stone: Right? So every single tack that you see here is workflows that I need her to do. And I can evoke it, but I can also say, like, if I say, research this, like, try to figure it out. Because, like, sometimes I’m on my phone, I really don’t care, and it doesn’t trigger all the time, that’s why I did this syntax.
267 00:27:01.550 ⇒ 00:27:17.599 Clarence Stone: mode definitions, right? Like, this is what she does for each one of these things. Like, presentations, slide decks, data analysis, edge finding, right? Evaluation of certain things, frameworks. So, like, for consulting, it’s like, hey, we have a problem.
268 00:27:17.600 ⇒ 00:27:27.179 Clarence Stone: Well, the first thing that we’re taught to do is find a proven framework to solve that problem. What I do for Lola is I have an RSS feed that hydrates her with a bunch of frameworks.
269 00:27:27.940 ⇒ 00:27:45.860 Clarence Stone: Right? So she’ll go, okay, within the frameworks that I have, right, and after searching online, this is the best framework that you should use. Well then, after that, like, she explains why this framework fits, and references the Notion Knowledge Base, and show the framework applied, not just describe.
270 00:27:47.240 ⇒ 00:27:48.210 Clarence Stone: Right.
271 00:27:49.860 ⇒ 00:28:06.069 Clarence Stone: If there’s no flag, try to figure it out, right? This is what it says here, right? The diagram of how she should behave, like, making, mermaid diagrams. This, this was a self-learned thing. You should always put this message in your agents.
272 00:28:06.570 ⇒ 00:28:08.500 Clarence Stone: I had an agent just delete
273 00:28:08.710 ⇒ 00:28:10.929 Clarence Stone: all these notes that I had.
274 00:28:13.240 ⇒ 00:28:24.890 Clarence Stone: knowledge base, right? So, okay, this is, this is a recursive loop, Utam, that I was talking about. So what happens is, when Lola finishes work, she creates a case study.
275 00:28:25.580 ⇒ 00:28:28.789 Clarence Stone: That case study gets reviewed by another agent.
276 00:28:29.490 ⇒ 00:28:34.239 Clarence Stone: And… and it gets feedback. I also leave feedback for it in Notion.
277 00:28:34.510 ⇒ 00:28:42.380 Clarence Stone: Right? And after that, that… the, the, like, the feedback from the methodology goes back into the engine.
278 00:28:42.980 ⇒ 00:28:48.439 Clarence Stone: say, it’s somewhere down here, in Methodology Ingester, where, like.
279 00:28:48.690 ⇒ 00:28:56.979 Clarence Stone: she’ll look online and say, oh, you should have learned this. Well, she’ll do it at 12 o’clock at night. She’ll search for the methodology she doesn’t have.
280 00:28:57.090 ⇒ 00:28:59.060 Clarence Stone: And add it to her library.
281 00:29:00.420 ⇒ 00:29:01.920 Clarence Stone: Based on my feedback.
282 00:29:03.900 ⇒ 00:29:19.659 Clarence Stone: Right? And then I have model routing instructions, how she should communicate. Primary communication, iMessage, right? Like, for your agents, it’s like probably Slack, right? Or email, right? You might want to put all of these… yeah, no Slack, no Trello, because…
283 00:29:19.660 ⇒ 00:29:31.259 Clarence Stone: because she’s a consulting agent, she tried to do that, like, naturally, right? Like, output standards, right? How should things be formatted? Yours is probably going to be much bigger than this.
284 00:29:32.120 ⇒ 00:29:33.040 Clarence Stone: Right.
285 00:29:33.230 ⇒ 00:29:35.769 Clarence Stone: So this is what a SOL file looks like.
286 00:29:36.910 ⇒ 00:29:51.629 Clarence Stone: I can share this with you. Mine’s not very organized. This is just, like, built along as I go, right? So what I would recommend is you create a format that fits for all three, and then say, hey, here’s everything that CSLs, EPs, and SLs do.
287 00:29:51.960 ⇒ 00:29:54.310 Clarence Stone: populate my, my soul files.
288 00:29:56.050 ⇒ 00:29:56.980 Clarence Stone: Right.
289 00:29:58.250 ⇒ 00:30:03.490 Clarence Stone: And then you can do an enrichment process saying, like, Greg is the CSO.
290 00:30:04.640 ⇒ 00:30:09.320 Clarence Stone: add Greg’s behaviors, into the CSO ND.
291 00:30:12.160 ⇒ 00:30:12.710 Uttam Kumaran: Yeah.
292 00:30:14.570 ⇒ 00:30:17.569 Clarence Stone: There’s very little of this that I’ve written myself.
293 00:30:18.510 ⇒ 00:30:23.560 Clarence Stone: This is probably the second time I’ve had to read this, and the first time was because things crashed.
294 00:30:26.790 ⇒ 00:30:36.829 Clarence Stone: Yeah. So, does that make sense? Like, we’re creating an agent proxy for each of the project roles. EP first, because it’s the most mechanical.
295 00:30:45.370 ⇒ 00:30:46.919 Uttam Kumaran: See, your brain’s gonna explode.
296 00:30:46.920 ⇒ 00:30:47.350 Clarence Stone: Don’t think…
297 00:30:47.350 ⇒ 00:30:51.110 Uttam Kumaran: don’t think too hard about it, but what I kind of want to… wanna…
298 00:30:51.340 ⇒ 00:31:03.959 Uttam Kumaran: align us back to is, like, these are just the foundations we need to do before we’re able to actually let AI start to run on its own. A lot of what the industry is doing now is finding out the hard way
299 00:31:04.130 ⇒ 00:31:10.859 Uttam Kumaran: That without these foundations, It’s really, really hard to, like, have agents sort of run on their own.
300 00:31:11.420 ⇒ 00:31:19.169 Uttam Kumaran: lucky, and especially within an organization. You know, so one thing that we need to figure out is to kind of crawl, walk, run here.
301 00:31:19.310 ⇒ 00:31:32.429 Uttam Kumaran: is how do we first use Erscher and agentic frameworks to, like, execute, and then how can we then hand some of that off to AI to do on its own? That’s gonna be, like, how to… kind of the process here that we… that we work through.
302 00:31:34.900 ⇒ 00:31:41.569 Clarence Stone: Yeah, and I think, you know, I think a success would be if you can just start with one, like.
303 00:31:41.570 ⇒ 00:31:56.159 Clarence Stone: EP question, right? Enrich, like, enrich the soul, just to target for that one use case, right? And then, make sure that you also create this tools file that gives them access and instructions on where they find these things.
304 00:31:56.260 ⇒ 00:32:13.410 Clarence Stone: Right? It’s like, hey, client folders, and this is your folder structure, this is where you find it, right? Or if you need to call, like, to external skills and processes, right? Like, use Notion for Notion MCP. Like, you have to give these agents permission to do these things, and tell them how it works.
305 00:32:13.410 ⇒ 00:32:18.099 Clarence Stone: Right? Like, diagramming tools, like, stuff like this. And it doesn’t have to be very detailed, clearly.
306 00:32:18.400 ⇒ 00:32:20.059 Clarence Stone: Quad wrote this by itself.
307 00:32:21.030 ⇒ 00:32:32.819 Clarence Stone: Right? And this is, like, here’s your local files that you use, right? This is your Notion databases, so next time I say, add it to your Notion, Lola knows where to put it. This is her Notion.
308 00:32:34.760 ⇒ 00:32:44.340 Clarence Stone: Right? Pre-built visuals, all of this, like, how should they do something, right? Souls should be, like, what are… who are they?
309 00:32:45.940 ⇒ 00:32:48.079 Clarence Stone: Right, and I would just start with one use case.
310 00:32:48.560 ⇒ 00:32:59.099 Clarence Stone: But, like, I think it begins with making sure you can get the data. And I realized that was the biggest mistake that I made in the beginning, is I didn’t know my services were crashing.
311 00:32:59.550 ⇒ 00:33:01.980 Clarence Stone: I didn’t know about my rate limit issues.
312 00:33:03.020 ⇒ 00:33:12.629 Clarence Stone: I had to build so much tooling to figure it out, and I realized I didn’t have to build that much, because there’s something called OpenTelemetry, you can just feed your data, and then you just have to wrap a dashboard around it.
313 00:33:15.230 ⇒ 00:33:21.040 Clarence Stone: Like, I was building one, like, logging feature at a time. It was just stupid.
314 00:33:23.750 ⇒ 00:33:35.069 Clarence Stone: Right? Like, some… there’s some really good safeguards that I can share with you. Like, this… this is really good, right? Autosave behavior, in case sessions crash, right? Stuff like that.
315 00:33:35.130 ⇒ 00:33:48.080 Clarence Stone: like, I think the biggest issue that you’re gonna face, when you do get these things up and running is that, like, if you aren’t alerted when these MCP connections break, when you run the cron drops.
316 00:33:48.460 ⇒ 00:33:49.370 Clarence Stone: like…
317 00:33:49.680 ⇒ 00:33:54.359 Clarence Stone: you’re not gonna know that it didn’t happen. And then you’re gonna wake up, and it didn’t do its job.
318 00:33:55.430 ⇒ 00:34:14.550 Uttam Kumaran: But see, some of this I’m also, like, I don’t need to hand it to an agent, like, we will just write the way we usually do it. Like, a meeting finished, write the thing, write the thing, write the thing. So, like, I’m not as worried about these because, again, I think you’re coming at it from an individual standpoint. Like, we’ll build a system around it, like, we’ll build logging and, like.
319 00:34:14.730 ⇒ 00:34:18.920 Uttam Kumaran: We won’t… we’ll run through APIs for everything, like, we won’t use MCPs.
320 00:34:19.159 ⇒ 00:34:27.759 Clarence Stone: Yeah, so what happened with the two weeks that were missing for the update problem then? Like, why didn’t we have logs for the calls?
321 00:34:28.389 ⇒ 00:34:29.859 Uttam Kumaran: Well, I’m gonna have to go ask.
322 00:34:30.090 ⇒ 00:34:34.310 Clarence Stone: Right? See, like, that’s what I’m saying, like, just, just, like, having a system where, like…
323 00:34:34.310 ⇒ 00:34:41.540 Uttam Kumaran: This is not an AI, but see, this is where I’m just, like, I just disagree, like, this is a… this is too… this is too fragile. This is not, like…
324 00:34:41.840 ⇒ 00:34:56.320 Uttam Kumaran: this is not a production system that you can run a company on. Like, this is a personal MCP-driven system. Like, yes, it’s finicky, but, like, we have APIs and pipelines that are landing
325 00:34:56.760 ⇒ 00:35:00.979 Uttam Kumaran: Transcripts into the repo, because we’re… we have hundreds of meetings a day.
326 00:35:01.280 ⇒ 00:35:02.870 Uttam Kumaran: That’s what I think, like, is the.
327 00:35:02.870 ⇒ 00:35:11.760 Clarence Stone: Yeah, and that’s fine, like, so don’t put the MCP instructions in, right? Like, the implementation’s still the same, you still need a SOL file and a tool.
328 00:35:11.760 ⇒ 00:35:14.319 Uttam Kumaran: Totally, totally. But I guess what I’m saying is, like.
329 00:35:14.870 ⇒ 00:35:18.590 Uttam Kumaran: This, I actually, I have no… I’m not worried about this.
330 00:35:19.020 ⇒ 00:35:20.839 Uttam Kumaran: as much as I’m worried about
331 00:35:21.010 ⇒ 00:35:27.190 Uttam Kumaran: having the data and having the APIs and endpoints. So, because if B is able to
332 00:35:27.400 ⇒ 00:35:38.239 Uttam Kumaran: complete his tasks as an EP every week in less than an hour, it is so trivial for me to hand that to an agent. What is not trivial is getting to that point, where B can do it.
333 00:35:38.780 ⇒ 00:35:46.079 Clarence Stone: Gotcha. Do you see what I mean? So the point that you’re saying, the fragile part, is automating it. Like, don’t do the cron stuff.
334 00:35:47.490 ⇒ 00:35:57.679 Uttam Kumaran: Yeah, well, I guess what I’m saying is I am not as worried about handing it to an agent as I am to handing it to a person to just show that they can do it fast.
335 00:35:57.960 ⇒ 00:36:03.829 Uttam Kumaran: I already know the Asian can do it faster and easier, but, like, if the human can’t do it.
336 00:36:04.070 ⇒ 00:36:07.669 Uttam Kumaran: The human is actually the litmus test for if the agent can do it.
337 00:36:08.030 ⇒ 00:36:11.350 Clarence Stone: That’s true. And right now, a human at my company cannot do this.
338 00:36:12.260 ⇒ 00:36:19.670 Uttam Kumaran: Because the agent is smarter. But the agent, like, the human actually requires the most context.
339 00:36:19.740 ⇒ 00:36:38.600 Uttam Kumaran: Right? So, like, if we solve for the human… for anybody at… like, this is where I actually think, B, you’re a great use case, you’re new, so you will require, like, okay, everything has to be written down, I need every SOP. When you’re ultimately able to run EP for a project, end-to-end, in less than an hour a week.
340 00:36:39.270 ⇒ 00:36:44.110 Uttam Kumaran: In 2 days, I will… We will have an agent start doing that.
341 00:36:44.790 ⇒ 00:36:47.630 Uttam Kumaran: But the… we can’t… until we get that.
342 00:36:48.060 ⇒ 00:36:52.929 Uttam Kumaran: Getting to hear, like, handing this to an agent will… it’ll… it’ll just mess up, you know?
343 00:36:52.930 ⇒ 00:37:09.959 Clarence Stone: Hang on, like, I think there’s a miscommunication here. Like, the only automated workflow that Lola does is self-enrichment. Like, I trigger all my research. I don’t say, like, hey, like, research this all the time. So, like, Lola is still very much an operated tool with me as the operator.
344 00:37:10.750 ⇒ 00:37:16.899 Uttam Kumaran: See, then I think that… I guess that’s what I missed, because I think the distance between Lola and, like, a Kircher command…
345 00:37:17.170 ⇒ 00:37:23.039 Uttam Kumaran: being like, run your EP… like, for B to wake up and be like, run my EP tasks.
346 00:37:25.250 ⇒ 00:37:30.670 Uttam Kumaran: I feel like we’re probably 2 weeks away from that, and then probably another 2 weeks from…
347 00:37:30.850 ⇒ 00:37:37.569 Uttam Kumaran: an EP agent at 9am, basically being like, I should run my EP test today.
348 00:37:37.570 ⇒ 00:37:41.089 Clarence Stone: Yeah, because it’s not any different. You layer the crown on top, right?
349 00:37:41.090 ⇒ 00:37:43.969 Uttam Kumaran: Yeah, but you’re right in that I’m coming the other way.
350 00:37:43.970 ⇒ 00:37:54.560 Clarence Stone: Exactly, I told you, like, we’re coming at this from completely two different directions, but it, like, I don’t have all the data he has, right? I have the agents.
351 00:37:54.560 ⇒ 00:38:04.949 Uttam Kumaran: Yeah, yeah, yeah, yeah. But, like, the thing is, like, what ClaudeBot is showing me is that this, like, ambient or autonomous agent system
352 00:38:05.060 ⇒ 00:38:08.320 Uttam Kumaran: It’s totally possible for very cheap.
353 00:38:08.630 ⇒ 00:38:10.750 Uttam Kumaran: At a company like ours.
354 00:38:11.560 ⇒ 00:38:19.499 Uttam Kumaran: Which shows me that, okay, like, as long as we keep nailing the foundations, and we get the humans to do it, it’ll be a pop, skip, and a jump.
355 00:38:19.860 ⇒ 00:38:20.919 Uttam Kumaran: Handing it to the.
356 00:38:20.920 ⇒ 00:38:31.230 Clarence Stone: honest, the only automation I would start with, if there was ever one, was… is creating a reviewer agent that clones feedback that UTAM has given.
357 00:38:33.060 ⇒ 00:38:37.039 Uttam Kumaran: Yeah, I mean, if I can give you a light version of this already.
358 00:38:37.040 ⇒ 00:38:39.010 Clarence Stone: He already has it written down.
359 00:38:39.010 ⇒ 00:38:44.219 Uttam Kumaran: Well, like, for all of our platform PRs, I mean, this is really, like, Sam is kind.
360 00:38:44.220 ⇒ 00:38:54.300 Clarence Stone: Because I created an agent that was how I review things, and that was the first automation I made, because I just don’t want to have to tell the same thing over and over again.
361 00:38:55.410 ⇒ 00:38:56.479 Uttam Kumaran: Yeah, so let me give you an.
362 00:38:56.480 ⇒ 00:39:01.169 Clarence Stone: And that’s the only one that matters, to be honest. Everything else, we don’t need to automate.
363 00:39:01.170 ⇒ 00:39:07.509 Uttam Kumaran: Let me give you an example of something that I did. So, I wanted to, I wanted to…
364 00:39:07.870 ⇒ 00:39:15.110 Uttam Kumaran: add KimmyK2 and 2.5, as an option,
365 00:39:15.430 ⇒ 00:39:20.889 Uttam Kumaran: in our platform for people to use. So I shipped this PR, using Codex.
366 00:39:21.080 ⇒ 00:39:28.140 Uttam Kumaran: We have a… we have a GitHub action that uses another agent to risk profile the PR,
367 00:39:28.730 ⇒ 00:39:31.599 Uttam Kumaran: and to auto-label it. And we used…
368 00:39:31.880 ⇒ 00:39:35.019 Uttam Kumaran: We use cooking terms, because I… I like to cook.
369 00:39:35.320 ⇒ 00:39:39.419 Uttam Kumaran: And so it’s, it’s, it’s, raw, half-baked, fully baked.
370 00:39:40.480 ⇒ 00:39:44.109 Uttam Kumaran: Low priority, high priority, and what is the effort to review it?
371 00:39:44.800 ⇒ 00:39:46.469 Uttam Kumaran: And is it a feature or a bug?
372 00:39:46.770 ⇒ 00:39:56.650 Uttam Kumaran: So what does this help us with? When Sam wakes up in the morning, and I’ve shipped, like, 30 PRs on, like, random things, he can be like, cool, anything half-baked, like, I’m just not gonna look at, because…
373 00:39:56.980 ⇒ 00:40:00.580 Uttam Kumaran: it’s still in progress, it needs to be worked on. Anything fully baked.
374 00:40:00.980 ⇒ 00:40:08.300 Uttam Kumaran: Okay, he can start to look at, and then ideally, anything… he should start with, anything fully baked, low review effort, he can probably look at for 10 seconds.
375 00:40:08.680 ⇒ 00:40:13.220 Uttam Kumaran: The other thing is an auto… it automatically goes through and tries to generate
376 00:40:13.510 ⇒ 00:40:21.650 Uttam Kumaran: descriptions, models, and it runs through a PR reviewer. So, we have AI automatically do its… do a first pass.
377 00:40:22.640 ⇒ 00:40:32.560 Uttam Kumaran: Is there tests? Are there security concerns? What is the focus areas for the reviewer? Right? And then we have a few other things that do the review, and then we sort of iterate.
378 00:40:32.730 ⇒ 00:40:37.000 Uttam Kumaran: Because what I told Sam is the limiting thing is gonna be there’s not enough Sam’s at my company.
379 00:40:37.690 ⇒ 00:40:39.319 Uttam Kumaran: But there needs to be a SAM.
380 00:40:39.520 ⇒ 00:40:42.559 Uttam Kumaran: And there needs to be a me. Someone needs to review everything that goes in.
381 00:40:43.520 ⇒ 00:40:53.900 Uttam Kumaran: So the limiting factor is not the fact that, like, we’re… these PR… like, everybody in the company’s gonna start shipping code to the platform. So I said, Sam, if I’m already shipping 30 PRs in, like.
382 00:40:54.110 ⇒ 00:40:55.240 Uttam Kumaran: a weekend?
383 00:40:56.020 ⇒ 00:40:59.890 Uttam Kumaran: Think about when we have 300 PRs, like, a week.
384 00:41:01.010 ⇒ 00:41:01.870 Uttam Kumaran: what…
385 00:41:02.160 ⇒ 00:41:11.430 Uttam Kumaran: happens. And we both are like, oh shit, like, the review process becomes a limiting factor. Okay, so we need to get to a point where
386 00:41:11.810 ⇒ 00:41:16.039 Uttam Kumaran: we can use AI, and we use ourselves, to start to auto…
387 00:41:16.350 ⇒ 00:41:18.810 Uttam Kumaran: Get AI to auto-review the easy stuff.
388 00:41:19.250 ⇒ 00:41:21.569 Uttam Kumaran: So that it can auto-push the easy stuff.
389 00:41:21.750 ⇒ 00:41:28.189 Uttam Kumaran: And then only give us the things that are… Either huge changes, Security flaws?
390 00:41:28.360 ⇒ 00:41:31.859 Uttam Kumaran: Or, like, Things where we just have subject matter expertise.
391 00:41:32.710 ⇒ 00:41:37.659 Uttam Kumaran: And, like, that’s… so we’ve already kind of built this system in the platform, because I said, Sam, I’m the guinea pig.
392 00:41:37.890 ⇒ 00:41:46.470 Uttam Kumaran: I am, like… I’m gonna try and really break every process by shipping, like, tons of features. Like, I went on Codex the other day, and I just…
393 00:41:46.570 ⇒ 00:41:55.110 Uttam Kumaran: I took 10 PR… 10 linear tickets we wrote 2 years ago about AI stuff I wanted to build, and I just said, go build each of these 10 things. I kicked them all off.
394 00:41:55.350 ⇒ 00:42:03.030 Uttam Kumaran: Right? And I was just like, fuck it. And then… and so I… what I… what I… what I had a really fun conversation with Sam with is I’m like, dude.
395 00:42:03.150 ⇒ 00:42:07.419 Uttam Kumaran: I don’t know, Clarence, do you listen to this Elon, thing on,
396 00:42:07.860 ⇒ 00:42:09.689 Uttam Kumaran: With the Stripe guy, the podcast?
397 00:42:09.690 ⇒ 00:42:10.720 Clarence Stone: Oh, yeah, yeah.
398 00:42:11.470 ⇒ 00:42:22.580 Uttam Kumaran: One thing that he loves to say, he’s like, I find the limiting factor, and I attack that. And so I told Sam, I said, dude, it’s not… we’re gonna quickly… people are gonna be able to code things and add to the platform.
399 00:42:22.740 ⇒ 00:42:25.200 Uttam Kumaran: But there’s not gonna be more reviewers at this company.
400 00:42:25.790 ⇒ 00:42:34.249 Uttam Kumaran: So, you need to find a way for the average PR… for the average codec session to actually push a PR that is, like, 90% of the way there.
401 00:42:35.130 ⇒ 00:42:38.150 Uttam Kumaran: And, for all PRs, we need to get better at
402 00:42:38.330 ⇒ 00:42:52.299 Uttam Kumaran: figuring out how to review them faster, if at all. And I said, those are your two objectives. So one thing he’s working on is, like, how do I build a better cursor environment, codex environment, so it has all the end keys, it could do end-to-end deployment and testing, so that’s what he’s working on.
403 00:42:53.320 ⇒ 00:43:04.430 Uttam Kumaran: And then we’re gonna… once we nail that, so, like, you could go in, basically, and it’ll… it’ll… you could ship a feature, we’re then gonna work more on, like, how do we review things faster, and have AI review half the stuff.
404 00:43:06.880 ⇒ 00:43:15.050 Uttam Kumaran: And so, I… what I’m… what we’re thinking about, B, is, like, how do… I want to do this very similarly at the… on the EP side, you know?
405 00:43:15.730 ⇒ 00:43:21.259 Uttam Kumaran: So that ultimately, you could think about a system where I can… I’m a CSO, and I can say.
406 00:43:21.620 ⇒ 00:43:22.910 Uttam Kumaran: Hey, EP.
407 00:43:23.190 ⇒ 00:43:26.799 Uttam Kumaran: Like, where are we at with… with our tasks for the week?
408 00:43:27.080 ⇒ 00:43:29.100 Uttam Kumaran: And it’s like, great.
409 00:43:29.260 ⇒ 00:43:34.830 Uttam Kumaran: I went ahead, I saw that there were 3 meetings earlier, I updated the Gantt, here it is for your review.
410 00:43:35.010 ⇒ 00:43:39.100 Uttam Kumaran: I updated linear tickets. Here are some updates that I was nervous about.
411 00:43:39.360 ⇒ 00:43:41.570 Uttam Kumaran: I’ve drafted these 4 emails.
412 00:43:42.140 ⇒ 00:43:44.759 Uttam Kumaran: Right? I drafted 3 slacks.
413 00:43:46.310 ⇒ 00:43:48.339 Uttam Kumaran: And let me know how else I can be helpful.
414 00:43:49.480 ⇒ 00:43:51.250 Uttam Kumaran: That is the dream.
415 00:43:52.320 ⇒ 00:43:54.989 Uttam Kumaran: However fast we can get there.
416 00:43:55.480 ⇒ 00:44:00.070 Uttam Kumaran: is the goal. And that is just the EP context, you know?
417 00:44:00.450 ⇒ 00:44:03.850 Uttam Kumaran: Because the CSO context is gonna be a bit harder, because
418 00:44:04.260 ⇒ 00:44:13.510 Uttam Kumaran: CSOs, ultimately, are the ones interacting with the clients. That is not something I’m interested in automating at this time. I actually want them to spend more time with clients.
419 00:44:13.850 ⇒ 00:44:19.549 Uttam Kumaran: Everybody needs to be in 8 hours of meetings with the clients, and it’s working with more clients per person.
420 00:44:19.710 ⇒ 00:44:20.500 Uttam Kumaran: Right?
421 00:44:20.850 ⇒ 00:44:23.949 Uttam Kumaran: Yeah. Because that… you can’t automate the human connection.
422 00:44:24.480 ⇒ 00:44:25.410 Brylle Girang: Right now.
423 00:44:25.690 ⇒ 00:44:26.300 Uttam Kumaran: So…
424 00:44:26.300 ⇒ 00:44:30.150 Clarence Stone: I think the best tools for CSOs are, like.
425 00:44:30.160 ⇒ 00:44:48.210 Clarence Stone: status updates, like, interconnects where they can see what’s going on, because if a client asks a question, they’re gonna have to, like, look through a bunch of things, ping a few people, like, are you done with it? Are you stuck? What’s happening? Right? Like, I… I want to be able to just pick up my phone, like I showed you with my open claw, and say, hey, what’s happening with this thing?
426 00:44:49.180 ⇒ 00:45:03.449 Clarence Stone: Right? Oh, Utam just pushed, the final, thing. It’s being reviewed. Hey, client, we just got it done. We’re just doing our final checks, making sure everything’s okay. You know, it’ll be ready on our weekly sync.
427 00:45:04.610 ⇒ 00:45:10.360 Clarence Stone: Right? Like, that looks so much more professional than saying, let me get back to you.
428 00:45:11.490 ⇒ 00:45:12.430 Clarence Stone: Right.
429 00:45:17.370 ⇒ 00:45:22.870 Clarence Stone: So, yeah, I think just focus on the EP, because, like, the others get a little bit complicated.
430 00:45:22.870 ⇒ 00:45:25.879 Uttam Kumaran: And… but then can I… I just… now that we’ve had this sort of, like…
431 00:45:26.580 ⇒ 00:45:31.779 Uttam Kumaran: North Star. I don’t know… I don’t know whether we are 6 weeks or 6 months.
432 00:45:32.390 ⇒ 00:45:35.500 Uttam Kumaran: From this future, but we are on the track.
433 00:45:35.700 ⇒ 00:45:41.990 Uttam Kumaran: However, to realign SB, and then I think this… we can end this call, I just need to get out of the house, go get a coffee.
434 00:45:42.370 ⇒ 00:45:45.160 Uttam Kumaran: This is where we are at as a company right now.
435 00:45:46.250 ⇒ 00:45:47.790 Uttam Kumaran: like, Luke…
436 00:45:48.150 ⇒ 00:45:55.719 Uttam Kumaran: sent me this message, any contacts you can give me on superposition. Connected with David on LinkedIn, and I’m gonna send up time to catch up with him.
437 00:45:57.790 ⇒ 00:45:59.380 Uttam Kumaran: This is what I said to Cursor.
438 00:45:59.580 ⇒ 00:46:01.270 Uttam Kumaran: Luke asked me this.
439 00:46:03.670 ⇒ 00:46:05.690 Uttam Kumaran: Superposition, what it is.
440 00:46:05.900 ⇒ 00:46:13.090 Uttam Kumaran: What’s our partnership with them? What we’ve done so far? Why David wants to meet Luke? Open threads David cares about.
441 00:46:13.820 ⇒ 00:46:20.590 Uttam Kumaran: I then took a screenshot, and I didn’t give him the answer. The nice thing I should have done is copied and pasted this.
442 00:46:20.810 ⇒ 00:46:25.220 Uttam Kumaran: because I’m… kind of rude, I literally sent the top part.
443 00:46:27.490 ⇒ 00:46:29.059 Uttam Kumaran: Luke asked me this.
444 00:46:30.220 ⇒ 00:46:31.900 Uttam Kumaran: And then guess what Luke said?
445 00:46:33.310 ⇒ 00:46:35.580 Uttam Kumaran: Oh, I didn’t even think about that.
446 00:46:37.980 ⇒ 00:46:42.249 Uttam Kumaran: On one hand, I’m happy people are learning.
447 00:46:43.120 ⇒ 00:46:44.690 Uttam Kumaran: On another hand…
448 00:46:44.990 ⇒ 00:46:52.460 Uttam Kumaran: I wanna punch my screen, like, I wanna… I wanna, like, punch my… I wanna, like, jump through my screen right now. It’s, like, so frustrating.
449 00:46:52.570 ⇒ 00:47:00.390 Uttam Kumaran: Right? So… I think this is maybe a good place to, like, put a bow on things, but…
450 00:47:01.130 ⇒ 00:47:13.880 Uttam Kumaran: For me, personally, I’m very excited that you’re here, I’m very excited that you’re, like, going really headfirst into this. I think let’s maybe think about planning tomorrow a little bit. I think maybe let’s try to… let’s try to chat when you’re up, you know, when we’re both up.
451 00:47:16.050 ⇒ 00:47:26.209 Uttam Kumaran: I have… one other thing that I’m gonna try to do, is that I… I used to do this, and maybe… I don’t know, Clarence, if you… like, I’m kind of interested in your meta.
452 00:47:26.460 ⇒ 00:47:31.270 Uttam Kumaran: thinking on, like, how to best… how we can all work together, but I kept an action log.
453 00:47:32.520 ⇒ 00:47:34.210 Uttam Kumaran: Like, in September?
454 00:47:34.890 ⇒ 00:47:38.280 Uttam Kumaran: Where I would wake up, and I would talk to ChatGPT, and basically…
455 00:47:38.460 ⇒ 00:47:42.690 Uttam Kumaran: I have a… I created a Chief of Staff Custom GPT, this is when custom GPTs were…
456 00:47:43.080 ⇒ 00:47:48.200 Uttam Kumaran: a thing, and I basically would say, here are all the things I have to do today.
457 00:47:48.520 ⇒ 00:47:49.240 Clarence Stone: Yep.
458 00:47:49.240 ⇒ 00:47:52.469 Uttam Kumaran: And I would talk for, like, 5 minutes.
459 00:47:52.660 ⇒ 00:47:54.470 Uttam Kumaran: It would then give me this output.
460 00:47:54.770 ⇒ 00:47:58.799 Uttam Kumaran: Give me all the tasks, what are the deadlines, and then it would give me, like, some of these.
461 00:47:59.480 ⇒ 00:48:00.170 Clarence Stone: Yeah.
462 00:48:01.080 ⇒ 00:48:09.260 Uttam Kumaran: My ask is, what is help… what do we think would be helpful for B that I did, given my li- given my limited time to, like.
463 00:48:09.410 ⇒ 00:48:18.039 Uttam Kumaran: If I could, I would just spend all day on the phone with you, and we’d just knock everything out. Unfortunately, I’m still in, like, 4 to 6 hours of meetings. What is helpful for me to do when I wake up?
464 00:48:18.160 ⇒ 00:48:25.329 Uttam Kumaran: to enable B, to just, like, accelerate. Like… Is my question.
465 00:48:29.250 ⇒ 00:48:34.070 Uttam Kumaran: like… Yeah, maybe just take a sec to think.
466 00:48:38.350 ⇒ 00:48:39.579 Clarence Stone: I think,
467 00:48:40.390 ⇒ 00:48:49.550 Clarence Stone: My opinion of… I made this statement, probably the second or third day, just like, you’re experiencing me when I… when I join the team here.
468 00:48:49.780 ⇒ 00:48:55.859 Clarence Stone: I said that… This organization is the one with the richest context.
469 00:48:56.410 ⇒ 00:48:57.930 Clarence Stone: But it’s everywhere.
470 00:48:59.730 ⇒ 00:49:08.750 Clarence Stone: Right. So, I think… I don’t… I think we just spammed a bunch of context at you, and that’s why it feels a little overwhelming, right?
471 00:49:09.030 ⇒ 00:49:22.170 Clarence Stone: I… I don’t think there’s anything else you can do, Utam. I think what B needs to do is just, like, digest that, and then say, this is, like, the context chain that I would link together to create, like, this one EP feature.
472 00:49:22.460 ⇒ 00:49:25.579 Clarence Stone: Right? And this is how I would plan future ones.
473 00:49:26.200 ⇒ 00:49:30.649 Clarence Stone: Right? And then, from there, we can figure it out, right?
474 00:49:30.650 ⇒ 00:49:31.780 Uttam Kumaran: Go step by step.
475 00:49:31.780 ⇒ 00:49:39.059 Clarence Stone: Yeah. Because, like, once you get the first one, dude, you’re gonna be off to… you’re gonna keep doing it until it’s done. Because it’s gonna be the same pattern.
476 00:49:40.200 ⇒ 00:49:41.060 Clarence Stone: Right?
477 00:49:41.340 ⇒ 00:49:52.069 Clarence Stone: And then, like, as you’re doing it, you’ll catch the errors, you’re gonna update the files and be like, don’t delete things, right? Stuff like that, that I learned the hardest way. So, yeah. So you guys have to start with the first one.
478 00:49:52.220 ⇒ 00:49:57.349 Uttam Kumaran: So today, what do we do? We did slide… we did decoration, and we did case study generation.
479 00:49:57.760 ⇒ 00:49:58.450 Brylle Girang: Yeah.
480 00:49:58.450 ⇒ 00:50:01.869 Uttam Kumaran: So I think maybe what’s helpful, Bea, is…
481 00:50:02.130 ⇒ 00:50:06.059 Uttam Kumaran: to use Kirscher, and to use your knowledge of the Forging the Future doc.
482 00:50:06.290 ⇒ 00:50:14.040 Uttam Kumaran: to create a list of all the other EP responsibilities, And let’s just… One by one.
483 00:50:14.510 ⇒ 00:50:15.369 Uttam Kumaran: Crush them out.
484 00:50:16.120 ⇒ 00:50:17.840 Brylle Girang: Yeah, I agree, I agree.
485 00:50:17.840 ⇒ 00:50:26.519 Uttam Kumaran: And I think what’s helpful, you know, one thing you said earlier is you’re like, I think that eliminating this task, like the slide creation, could save an hour.
486 00:50:26.840 ⇒ 00:50:28.040 Uttam Kumaran: That is what…
487 00:50:28.240 ⇒ 00:50:35.689 Uttam Kumaran: your billboard is going to be for why… for your impact on the company. So if you can save those.
488 00:50:35.930 ⇒ 00:50:41.750 Uttam Kumaran: Right? Like, for example, you just went through slide creation today. If you’re like, hey, this took me 90 minutes.
489 00:50:42.470 ⇒ 00:50:45.809 Uttam Kumaran: And we will work to take it down to 15.
490 00:50:46.370 ⇒ 00:50:51.669 Uttam Kumaran: you’re gonna come to me and be like, hey, I just took this task that everybody is doing.
491 00:50:51.800 ⇒ 00:50:53.730 Uttam Kumaran: One to twi- two times a week.
492 00:50:53.840 ⇒ 00:50:56.609 Uttam Kumaran: From an hour and a half to 15 minutes.
493 00:50:56.740 ⇒ 00:51:02.229 Uttam Kumaran: Right? That is an hour 15 at our average salary. Here is how much that saves us.
494 00:51:03.290 ⇒ 00:51:07.130 Uttam Kumaran: That is, like, the poster. That is the billboard.
495 00:51:07.250 ⇒ 00:51:13.379 Uttam Kumaran: So, for the EP, I think Clarence is right, is that let’s go task by task.
496 00:51:14.240 ⇒ 00:51:19.490 Uttam Kumaran: And see if we can… If you want to create that task list, I can help you stack rank.
497 00:51:19.660 ⇒ 00:51:28.289 Uttam Kumaran: Ideally, we could double by the task that, like, I… because I have… immediately need some help with things, so we can match those, you can do them using AI.
498 00:51:28.630 ⇒ 00:51:34.140 Uttam Kumaran: I think, Clarence, probably where we can start keep discussing in Slack is, like, versioning?
499 00:51:34.740 ⇒ 00:51:40.160 Uttam Kumaran: Because… I think for each of these tasks, we’re gonna have, like, A proof of concept.
500 00:51:40.470 ⇒ 00:51:43.590 Uttam Kumaran: an MVP, And, like, a V1.
501 00:51:43.860 ⇒ 00:51:50.579 Uttam Kumaran: what I don’t want us to do is, like, think about taking something from an hour and a half to 15, and then 15 to 5. 15 to 5 is… is…
502 00:51:50.620 ⇒ 00:51:52.350 Clarence Stone: Irrelevant to me right now.
503 00:51:52.730 ⇒ 00:51:56.640 Uttam Kumaran: I’m thinking about anything that’s taking an hour, can we get to 15.
504 00:51:56.890 ⇒ 00:51:57.510 Clarence Stone: Yep.
505 00:51:57.510 ⇒ 00:51:58.879 Uttam Kumaran: Kind of following?
506 00:51:59.120 ⇒ 00:51:59.710 Clarence Stone: Yeah.
507 00:52:00.340 ⇒ 00:52:06.759 Uttam Kumaran: the mind and the brain and the urge is gonna try to think… tell you to take it from 15 to 5. Don’t…
508 00:52:06.930 ⇒ 00:52:07.970 Uttam Kumaran: Go into that.
509 00:52:09.030 ⇒ 00:52:17.640 Uttam Kumaran: I can tell Clarence is laughing, because he is a 15 to 5 type person. I am an hour to 15, we win. Let’s go to the next thing.
510 00:52:17.780 ⇒ 00:52:29.560 Uttam Kumaran: But this is… we… the three of us need to fight over this, and, like, we need to… this will be a nuanced conversation by task, but I’m telling you that, like, if we’re… some things, dude, people are not even doing.
511 00:52:30.180 ⇒ 00:52:36.959 Uttam Kumaran: Right? They’re not even doing Nomadic, because it takes 5 hours to do it, so they don’t do it. So part of this is, like, can all the things be done?
512 00:52:37.160 ⇒ 00:52:39.490 Uttam Kumaran: And then can all those be done in 15 minutes?
513 00:52:40.720 ⇒ 00:52:46.510 Uttam Kumaran: Then we’ll take a… we’ll take a breather, and this will probably be in, like, 2 weeks, 3 weeks. We’ll be like, okay.
514 00:52:47.530 ⇒ 00:52:51.110 Uttam Kumaran: What else? And then we’ll move… and then we’ll move into the next phase.
515 00:52:51.230 ⇒ 00:52:55.640 Uttam Kumaran: So, I feel like… That’s the mission I have in my brain.
516 00:52:56.310 ⇒ 00:53:01.379 Uttam Kumaran: I think ending this call, you could take this transcript and help you build a plan if you need.
517 00:53:01.530 ⇒ 00:53:09.239 Uttam Kumaran: But, yeah, I couldn’t be more excited about this. I would talk about it more, but, like, my… I literally need to get up and do something.
518 00:53:09.450 ⇒ 00:53:12.510 Uttam Kumaran: But this is really, really great. Like, I’m pumped for this.
519 00:53:13.850 ⇒ 00:53:30.809 Brylle Girang: I am too, I am too. Yeah, I think that that’s a good plan, starting with laying out the steps. Maybe we can do the same with how you laid out the PRs, tag them, how difficult they are, how long they take, and then let’s prioritize based on that.
520 00:53:31.250 ⇒ 00:53:34.340 Uttam Kumaran: Yeah, and if you have any question,
521 00:53:34.530 ⇒ 00:53:36.610 Uttam Kumaran: I’ll tell you, the only, like…
522 00:53:37.300 ⇒ 00:53:43.040 Uttam Kumaran: I’m gonna… I’m gonna say something probably more controversial in, like, some of the questions you have
523 00:53:43.280 ⇒ 00:53:47.639 Uttam Kumaran: You’re not gonna be able to get… you’re not gonna get answers from people that help you out.
524 00:53:48.290 ⇒ 00:53:52.410 Uttam Kumaran: So, meaning, like, you may go ask someone, how long does this take you?
525 00:53:53.830 ⇒ 00:53:56.020 Uttam Kumaran: It’s… it’s so subjective.
526 00:53:56.470 ⇒ 00:53:58.179 Brylle Girang: So instead, I think…
527 00:53:58.180 ⇒ 00:53:59.560 Uttam Kumaran: I’m a good source.
528 00:54:00.200 ⇒ 00:54:06.920 Uttam Kumaran: And Ricoh is probably a good source for authentically, like, how long does something take, and why?
529 00:54:08.240 ⇒ 00:54:15.920 Uttam Kumaran: And Clarence is probably a good source for, like, In a perfect world, like.
530 00:54:16.260 ⇒ 00:54:20.599 Uttam Kumaran: What it could look like, and it’s up to you to land us somewhere in the middle.
531 00:54:21.030 ⇒ 00:54:25.360 Clarence Stone: Right. I was his business OS Architect, so I’m gonna wanna wish that everything works perfectly.
532 00:54:25.360 ⇒ 00:54:25.920 Uttam Kumaran: Yeah.
533 00:54:26.910 ⇒ 00:54:30.619 Uttam Kumaran: And I’m gonna pull us towards, like, yo, I’m dying today.
534 00:54:31.170 ⇒ 00:54:36.860 Uttam Kumaran: And, like, your… your job is to land us somewhere in the middle, and so if you… if you create that list, I’ll help you prioritize.
535 00:54:38.230 ⇒ 00:54:39.040 Uttam Kumaran: then…
536 00:54:39.430 ⇒ 00:54:44.789 Uttam Kumaran: The next biggest thing after that is gonna be how do we then take those learnings and get it to every single EP.
537 00:54:46.160 ⇒ 00:54:49.819 Uttam Kumaran: But I think… Problem for when we get there, so…
538 00:54:49.820 ⇒ 00:54:50.410 Clarence Stone: Yeah.
539 00:54:51.100 ⇒ 00:54:51.630 Uttam Kumaran: Okay.
540 00:54:52.690 ⇒ 00:54:55.209 Clarence Stone: You guys got me energized, now I’m gonna build something.
541 00:54:55.380 ⇒ 00:55:02.650 Uttam Kumaran: I’m gonna go… I’m going to a coffee shop, I gotta… I’m gonna go get a double shot, I’m gonna… I gotta… yeah, I’m gonna go grab a couple more things up, for sure. Me too, I’m hyped up.
542 00:55:02.650 ⇒ 00:55:03.839 Clarence Stone: Do you use Super Whisper?
543 00:55:04.630 ⇒ 00:55:05.400 Clarence Stone: Or whisper.
544 00:55:05.400 ⇒ 00:55:06.070 Brylle Girang: Sorry?
545 00:55:06.070 ⇒ 00:55:08.220 Clarence Stone: Do you use Whisper? Because you were talking about.
546 00:55:08.220 ⇒ 00:55:08.620 Brylle Girang: Yeah.
547 00:55:08.710 ⇒ 00:55:10.010 Clarence Stone: Yeah, yeah. Okay, so you.
548 00:55:10.010 ⇒ 00:55:10.739 Brylle Girang: Yeah, I do, I do.
549 00:55:10.740 ⇒ 00:55:13.069 Clarence Stone: And it triggers the voice.
550 00:55:13.250 ⇒ 00:55:13.889 Clarence Stone: Oh, yeah.
551 00:55:13.890 ⇒ 00:55:14.230 Brylle Girang: Yeah.
552 00:55:14.230 ⇒ 00:55:23.800 Clarence Stone: I want to put a really small recursive model, and then put a hotkey, and then UTAM can put in any keyword or question, and it’ll parse all the MD files and just dump context.
553 00:55:24.270 ⇒ 00:55:25.200 Uttam Kumaran: What do you mean?
554 00:55:25.620 ⇒ 00:55:37.319 Clarence Stone: Like, you know that question that Luke asked you? Like, make a hotkey on your keyboard, start typing the thing, or you can even highlight it, and then put the hotkey in, and it’ll just… Well, you know, you know what we’re 2 days… you know what we’re 2 or 3 days from?
555 00:55:37.320 ⇒ 00:55:42.019 Uttam Kumaran: the Brainforge assistant having… Act context of everything and platform.
556 00:55:42.310 ⇒ 00:55:45.219 Clarence Stone: Yeah, so have it worth, and then dump it.
557 00:55:45.850 ⇒ 00:55:53.889 Uttam Kumaran: Well, let me tell you… let me tell you the roadmap on the… and the assistant, because I just nailed it down on Wednesday, because I called a friend of mine that’s building this.
558 00:55:55.560 ⇒ 00:56:02.150 Uttam Kumaran: by probably the weekend, I would… I… I… the way I would have come about it would have been, like, at Brainforge Assistant, help with this.
559 00:56:03.170 ⇒ 00:56:12.500 Uttam Kumaran: Probably in 2 weeks, Brain Forge Assistant will, in a thread that our group will have, say, hey, I can answer in this way, do you want me to answer?
560 00:56:13.570 ⇒ 00:56:14.550 Uttam Kumaran: and click approve.
561 00:56:15.950 ⇒ 00:56:22.229 Uttam Kumaran: And then, probably 6 months from now, it will respond to some of them on its own, like, no problem.
562 00:56:23.070 ⇒ 00:56:26.339 Uttam Kumaran: So… I think my take there is, like…
563 00:56:26.340 ⇒ 00:56:29.309 Clarence Stone: tool for myself. You guys don’t have to use it. I use it because.
564 00:56:29.310 ⇒ 00:56:34.680 Uttam Kumaran: No, no, no, no, no, I hear… but I guess I’m not… I’m not falling, you’re saying that, like, it just auto… it autofills?
565 00:56:35.370 ⇒ 00:56:35.990 Clarence Stone: Yeah.
566 00:56:36.650 ⇒ 00:56:42.380 Uttam Kumaran: But in Slack, I don’t know how… I don’t know if it could… oh, I don’t know how it could do that. Wherever your cursor is.
567 00:56:42.790 ⇒ 00:56:44.880 Clarence Stone: No, just like how WhisperSync works.
568 00:56:45.540 ⇒ 00:56:48.900 Clarence Stone: Oh, sorry, Super Whisper works, right? You hit a hot…
569 00:56:48.900 ⇒ 00:56:53.560 Uttam Kumaran: your highlighted cursor is, and then puts… does auto… so it’s like auto… it’s like auto… it’s like.
570 00:56:53.560 ⇒ 00:56:55.799 Clarence Stone: A lot of context. Yeah.
571 00:56:56.410 ⇒ 00:56:57.789 Uttam Kumaran: Wait, you have this built?
572 00:56:57.990 ⇒ 00:57:00.490 Clarence Stone: Yeah, I, like, agents are turning on it right now.
573 00:57:01.140 ⇒ 00:57:01.879 Uttam Kumaran: I’m okay, well, yeah.
574 00:57:01.880 ⇒ 00:57:02.679 Clarence Stone: our agents.
575 00:57:03.250 ⇒ 00:57:06.310 Uttam Kumaran: Okay, well, send me… send me a screenshot when it’s working.
576 00:57:06.310 ⇒ 00:57:10.900 Clarence Stone: I have to use, like, different AI things, I’m not gonna fucking explain it over and over again.
577 00:57:12.560 ⇒ 00:57:16.820 Clarence Stone: Right? Like, I just log the conversation to an MD and just dump the context.
578 00:57:17.550 ⇒ 00:57:18.370 Uttam Kumaran: Hmm.
579 00:57:20.620 ⇒ 00:57:23.409 Clarence Stone: This is part of a much bigger architecture, but…
580 00:57:23.860 ⇒ 00:57:24.900 Uttam Kumaran: Alright.
581 00:57:26.330 ⇒ 00:57:28.410 Uttam Kumaran: Okay, guys, this is great.
582 00:57:30.290 ⇒ 00:57:31.030 Uttam Kumaran: Cool.
583 00:57:31.350 ⇒ 00:57:32.380 Clarence Stone: Cool, awesome.
584 00:57:32.380 ⇒ 00:57:35.580 Uttam Kumaran: Clarence, I owe you EY stuff, I’m gonna go work on that now.
585 00:57:36.800 ⇒ 00:57:39.129 Clarence Stone: Oh, no, dude, no worries. Like.
586 00:57:39.740 ⇒ 00:57:41.250 Uttam Kumaran: Alright, I mean, that’s like…
587 00:57:41.250 ⇒ 00:57:48.220 Clarence Stone: when you get to the coffee shop, because I just posted a really good opus report that would, like, make you look like a fucking hero.
588 00:57:48.220 ⇒ 00:57:50.510 Uttam Kumaran: Okay, okay, okay, I’ll call you when I’m, when I’m back.
589 00:57:50.510 ⇒ 00:57:54.290 Clarence Stone: when you look… look at it. Okay, okay. Yeah. Alright.
590 00:57:54.290 ⇒ 00:57:56.440 Uttam Kumaran: Thank you, guys. Appreciate the time.
591 00:57:57.080 ⇒ 00:57:57.670 Clarence Stone: How are you?
592 00:57:58.280 ⇒ 00:57:59.400 Uttam Kumaran: Bye, talk to you soon.