Meeting Title: Sprint Review: AI Team Date: 2025-10-24 Meeting participants: Casie Aviles, Samuel Roberts, Mustafa Raja, Rico Rejoso, Uttam Kumaran
WEBVTT
1 00:01:35.290 ⇒ 00:01:36.420 Samuel Roberts: Hey, Casey.
2 00:01:39.520 ⇒ 00:01:40.950 Casie Aviles: Hey. Hey, Sean.
3 00:01:41.740 ⇒ 00:01:44.749 Samuel Roberts: So, Tom’s Eddie’s running a little behind,
4 00:01:45.170 ⇒ 00:01:46.499 Samuel Roberts: I don’t know if you saw…
5 00:01:46.840 ⇒ 00:01:47.740 Casie Aviles: Yeah…
6 00:01:48.140 ⇒ 00:01:52.769 Samuel Roberts: So I wasn’t sure if we were needing to bump this or not, but he didn’t respond yet, so I figured we’d hop on.
7 00:01:53.640 ⇒ 00:01:54.360 Mustafa Raja: Hang.
8 00:01:59.630 ⇒ 00:02:01.340 Samuel Roberts: How’s it going for you two? How was the week?
9 00:02:02.480 ⇒ 00:02:04.150 Mustafa Raja: I’m good.
10 00:02:05.580 ⇒ 00:02:06.930 Mustafa Raja: I think that’s good.
11 00:02:08.330 ⇒ 00:02:09.199 Samuel Roberts: It’s good.
12 00:02:09.590 ⇒ 00:02:10.270 Mustafa Raja: Yeah.
13 00:02:17.500 ⇒ 00:02:22.460 Samuel Roberts: Yeah, pretty much the same here, nothing… I’ve been crazy.
14 00:02:22.460 ⇒ 00:02:25.230 Mustafa Raja: Yeah, nothing crazy. Smooth.
15 00:02:32.330 ⇒ 00:02:34.619 Samuel Roberts: Man, it’s finally getting cold here.
16 00:02:35.330 ⇒ 00:02:36.090 Mustafa Raja: Oh.
17 00:02:36.750 ⇒ 00:02:39.670 Samuel Roberts: It’s, turning from, you know, it was…
18 00:02:39.860 ⇒ 00:02:42.890 Samuel Roberts: Still very warm for the summer, even into October.
19 00:02:42.990 ⇒ 00:02:47.150 Samuel Roberts: But it finally is getting… getting chilly here, and… I gotta…
20 00:02:47.460 ⇒ 00:02:52.989 Samuel Roberts: in my… my office here in the attic, I still had air conditioners in the windows, I gotta…
21 00:02:53.410 ⇒ 00:02:56.450 Samuel Roberts: Close those now, close those now, because I think it’s letting…
22 00:02:58.210 ⇒ 00:03:01.190 Mustafa Raja: That’s nice. How warm does it get in summer?
23 00:03:01.890 ⇒ 00:03:07.859 Samuel Roberts: Oh, man, so, it doesn’t get crazy hot, except for, like, maybe one week in August?
24 00:03:08.060 ⇒ 00:03:11.630 Samuel Roberts: But even into October, we were getting, like.
25 00:03:12.470 ⇒ 00:03:15.999 Samuel Roberts: We were against, like, 80 degree Fahrenheit days.
26 00:03:17.200 ⇒ 00:03:20.390 Mustafa Raja: Which is, like, what? Let me see what…
27 00:03:20.650 ⇒ 00:03:23.110 Samuel Roberts: 26, 27 degrees, yeah.
28 00:03:23.670 ⇒ 00:03:28.580 Samuel Roberts: So yeah, I probably was getting up to, like, 30 at the highest, but… 30C.
29 00:03:28.820 ⇒ 00:03:30.810 Samuel Roberts: I was trying to get better with,
30 00:03:31.030 ⇒ 00:03:35.250 Samuel Roberts: learning the difference between Celsius and Fahrenheit when I was living in London.
31 00:03:35.540 ⇒ 00:03:36.300 Mustafa Raja: Oh.
32 00:03:37.050 ⇒ 00:03:42.900 Samuel Roberts: Because everything, obviously, was Celsius there, but I was so used to Fahrenheit, and I got pretty good at…
33 00:03:43.640 ⇒ 00:03:46.309 Samuel Roberts: getting more comfortable using Celsius.
34 00:03:47.390 ⇒ 00:03:54.899 Samuel Roberts: Because I don’t want to just do… do the conversion, I want to be, like, able to say, like, oh, yeah, it’s warm, it’s like… and then have a number, and I was having a hard time with that.
35 00:03:55.110 ⇒ 00:03:58.229 Mustafa Raja: How long were you in London?
36 00:03:58.890 ⇒ 00:04:06.010 Samuel Roberts: We were there for 2 years, so we left, we were there from, like, October 21 to October 23.
37 00:04:06.510 ⇒ 00:04:07.210 Mustafa Raja: Nice.
38 00:04:07.540 ⇒ 00:04:08.390 Samuel Roberts: Yeah.
39 00:04:10.850 ⇒ 00:04:12.219 Samuel Roberts: Yeah, it was a fun time.
40 00:04:13.390 ⇒ 00:04:18.520 Samuel Roberts: But yeah, getting used to that wasn’t quite long enough for me to really… Really get used to it.
41 00:04:20.120 ⇒ 00:04:20.810 Mustafa Raja: Yeah.
42 00:04:22.640 ⇒ 00:04:23.320 Samuel Roberts: But…
43 00:04:24.780 ⇒ 00:04:28.009 Mustafa Raja: I think the major difference would have been,
44 00:04:28.230 ⇒ 00:04:31.420 Mustafa Raja: cars, they have left-hand ones, and US.
45 00:04:31.420 ⇒ 00:04:31.920 Samuel Roberts: Yeah.
46 00:04:31.920 ⇒ 00:04:37.589 Mustafa Raja: And one’s… Definitely. So do you need a new license for that, or how does that work?
47 00:04:37.820 ⇒ 00:04:43.289 Samuel Roberts: So, no, because we were only there… so, for… I forget what, like, a year, I think?
48 00:04:43.910 ⇒ 00:04:48.080 Samuel Roberts: You’re allowed to use your license, so basically if you’re visiting and stuff, you don’t need a new license.
49 00:04:48.670 ⇒ 00:04:49.430 Mustafa Raja: Okay.
50 00:04:49.630 ⇒ 00:04:51.060 Samuel Roberts: And honestly, like…
51 00:04:51.060 ⇒ 00:04:52.000 Mustafa Raja: confine.
52 00:04:52.400 ⇒ 00:04:59.150 Samuel Roberts: It’s very confusing. We didn’t drive a ton, you know, like, we were living in London, so the tube and everything was pretty good to take in the…
53 00:04:59.290 ⇒ 00:05:00.309 Samuel Roberts: The subway around.
54 00:05:00.830 ⇒ 00:05:02.880 Samuel Roberts: But the few times we did.
55 00:05:03.080 ⇒ 00:05:05.699 Samuel Roberts: It was… it was weird. It was weird.
56 00:05:06.840 ⇒ 00:05:12.740 Samuel Roberts: And, like, you’re going fine, and then every once in a while you forget, and you make, like, a turn into the wrong lane or something.
57 00:05:12.740 ⇒ 00:05:13.569 Mustafa Raja: Oh, yeah.
58 00:05:14.600 ⇒ 00:05:18.320 Samuel Roberts: And the weirdest part are the rotaries. The roundabouts.
59 00:05:18.600 ⇒ 00:05:19.270 Mustafa Raja: Oh.
60 00:05:19.700 ⇒ 00:05:22.609 Samuel Roberts: Because everything goes a different way around the circle.
61 00:05:23.070 ⇒ 00:05:26.389 Mustafa Raja: Yeah, we do have a lot of those over here.
62 00:05:26.680 ⇒ 00:05:31.200 Samuel Roberts: Yeah, well, see, I grew up with a lot in, like, New England, in Boston, there’s a lot.
63 00:05:31.550 ⇒ 00:05:48.810 Samuel Roberts: But then you come to the Midwest here, and there’s not as many, so people aren’t used to them, so my wife wasn’t nearly as used to them. But it didn’t matter, because like I said, everything goes the wrong way for me, in my head, so… in London, it didn’t matter. I still had to be like, as I’d drive, I would tell my wife, like, remind me I have to go this way, not that way, you know?
64 00:05:48.810 ⇒ 00:05:49.730 Mustafa Raja: Yeah.
65 00:05:52.360 ⇒ 00:05:53.270 Samuel Roberts: Hey, Rico.
66 00:05:55.700 ⇒ 00:05:56.640 Rico Rejoso: Hey guys.
67 00:05:57.020 ⇒ 00:05:57.690 Mustafa Raja: Right.
68 00:05:58.300 ⇒ 00:05:59.200 Samuel Roberts: How’s it going?
69 00:05:59.690 ⇒ 00:06:00.840 Rico Rejoso: Good morning.
70 00:06:01.160 ⇒ 00:06:02.940 Rico Rejoso: But it shows very in…
71 00:06:04.930 ⇒ 00:06:06.450 Samuel Roberts: Okay, no worries.
72 00:06:16.370 ⇒ 00:06:20.310 Samuel Roberts: So do we… do something with Langviews last night, is that what I saw?
73 00:06:20.310 ⇒ 00:06:28.730 Mustafa Raja: Yes, I think we found a ticket about consolidating, or what’s it called, system prompts.
74 00:06:28.910 ⇒ 00:06:32.710 Samuel Roberts: Okay. I remembered using LangFuse.
75 00:06:32.710 ⇒ 00:06:35.900 Mustafa Raja: For him to do it, and it was pretty good.
76 00:06:36.430 ⇒ 00:06:37.020 Samuel Roberts: Okay, cool.
77 00:06:37.020 ⇒ 00:06:38.719 Mustafa Raja: posted that tootham.
78 00:06:38.950 ⇒ 00:06:43.730 Mustafa Raja: And we just launched it for our internal stuff.
79 00:06:44.700 ⇒ 00:06:46.609 Samuel Roberts: Okay, cool, yeah, I saw that, I gotta…
80 00:06:46.610 ⇒ 00:06:51.770 Mustafa Raja: I have another one. The other was, something other than Langfuse.
81 00:06:52.150 ⇒ 00:06:55.320 Samuel Roberts: No, no, no, I mean, I had another LangFuse I had set up.
82 00:06:55.320 ⇒ 00:06:55.800 Mustafa Raja: In just a hot…
83 00:06:55.800 ⇒ 00:07:04.629 Samuel Roberts: mode, but I never used it. It was just I was testing it out. But I’ll… there’s nothing… I’m looking to see if there’s anything there. I don’t think there’s anything there, I just was testing it before.
84 00:07:04.630 ⇒ 00:07:05.030 Mustafa Raja: Okay.
85 00:07:05.030 ⇒ 00:07:08.400 Samuel Roberts: Because I was confused why I was invited to… .
86 00:07:09.840 ⇒ 00:07:10.970 Mustafa Raja: Yeah. Interluded.
87 00:07:11.030 ⇒ 00:07:11.680 Samuel Roberts: Wow.
88 00:07:11.680 ⇒ 00:07:14.880 Mustafa Raja: So, we changed the name of the.
89 00:07:14.880 ⇒ 00:07:20.670 Samuel Roberts: Yeah, I realize that now, when I saw the email, I was like, what? And then I saw that, yeah, okay, we’re good, we’re good.
90 00:07:20.670 ⇒ 00:07:35.240 Mustafa Raja: It’s now bringing forth. Utam was asking if we can, see usage for Maestra in it, and it does have observability, so I’m wondering if we can hook Maestra agents with it.
91 00:07:36.350 ⇒ 00:07:39.940 Samuel Roberts: Yeah, I think there’s some things we can do there. Monster seems a little better than, obviously, N8N.
92 00:07:40.370 ⇒ 00:07:41.549 Mustafa Raja: Yeah, yeah, yeah, I suppose.
93 00:07:41.550 ⇒ 00:07:48.769 Samuel Roberts: And, yeah, I think we should be able to get in there and get logs somewhere, you know, depending on where we want the logs, but…
94 00:07:48.770 ⇒ 00:07:53.339 Mustafa Raja: Yeah, so LangFuse has observability, do we want it there?
95 00:07:53.990 ⇒ 00:07:56.240 Samuel Roberts: I think it might make sense…
96 00:07:56.920 ⇒ 00:07:57.330 Mustafa Raja: Yeah.
97 00:07:58.470 ⇒ 00:07:59.080 Mustafa Raja: I mean, if you have…
98 00:07:59.080 ⇒ 00:08:01.559 Samuel Roberts: If we’re gonna… if we’re gonna use it, if we’re already gonna use it, yeah, exactly.
99 00:08:01.560 ⇒ 00:08:02.330 Mustafa Raja: Yeah.
100 00:08:04.740 ⇒ 00:08:05.320 Mustafa Raja: Yeah.
101 00:08:05.320 ⇒ 00:08:06.339 Samuel Roberts: Makes sense, right?
102 00:08:07.480 ⇒ 00:08:12.099 Samuel Roberts: Yeah, sorry, I’m just trying to delete that other organization, because I don’t need it anymore.
103 00:08:12.910 ⇒ 00:08:13.760 Mustafa Raja: Yeah.
104 00:08:16.270 ⇒ 00:08:20.839 Samuel Roberts: Okay. Yeah, I’ll have to dig a little bit, because I… let me monstra…
105 00:08:22.540 ⇒ 00:08:26.210 Mustafa Raja: I didn’t dig too deep into that, I just… .
106 00:08:26.210 ⇒ 00:08:26.680 Samuel Roberts: Yeah.
107 00:08:26.680 ⇒ 00:08:33.160 Mustafa Raja: the idea that came from with them when I looked into philanthes and saw that they have observability, they trace.
108 00:08:33.169 ⇒ 00:08:36.259 Samuel Roberts: Yeah, no, I’m finding, I’m finding a page right here.
109 00:08:36.419 ⇒ 00:08:40.409 Samuel Roberts: on… on Langfuse’s, docs.
110 00:08:41.490 ⇒ 00:08:41.919 Mustafa Raja: Hmm…
111 00:08:41.929 ⇒ 00:08:44.939 Samuel Roberts: I don’t remember existing before. When is this, when was this posted?
112 00:08:45.379 ⇒ 00:08:51.269 Samuel Roberts: Last updated October 2nd. Okay, so this might have been even before I was looking at it, so this might be new. Cool.
113 00:08:52.119 ⇒ 00:08:54.249 Samuel Roberts: Yeah, so we can set that up.
114 00:08:56.570 ⇒ 00:08:57.280 Mustafa Raja: Yeah.
115 00:09:04.050 ⇒ 00:09:05.070 Samuel Roberts: Yeah.
116 00:09:05.760 ⇒ 00:09:09.560 Samuel Roberts: Yeah, I tested a few things. LineFuse seemed alright, but I was having a hard time with…
117 00:09:09.710 ⇒ 00:09:11.980 Samuel Roberts: Copilot kit, but that was because…
118 00:09:12.130 ⇒ 00:09:18.010 Samuel Roberts: excuse me, the way Copilot Kit was working, it…
119 00:09:20.330 ⇒ 00:09:23.310 Samuel Roberts: So, Copilot Kit doesn’t actually have…
120 00:09:25.710 ⇒ 00:09:30.409 Samuel Roberts: Tracing, necessarily, because it’s not the same kind of, like, agent framework.
121 00:09:31.040 ⇒ 00:09:37.320 Samuel Roberts: So I think we would want to make sure to use… yeah, now that… and because we were using that on top of NAN, I was having a hard time…
122 00:09:37.580 ⇒ 00:09:42.839 Samuel Roberts: Getting it to, like, do the proper tracing.
123 00:09:44.230 ⇒ 00:09:46.150 Samuel Roberts: I was getting a lot of, like, extra stuff.
124 00:09:48.400 ⇒ 00:09:49.690 Mustafa Raja: Yeah.
125 00:09:49.900 ⇒ 00:09:56.279 Samuel Roberts: But now that we’re adding Mastra, we can probably do it, and it’ll go right into Mastra, or from Mastra into Langfeas, I should say.
126 00:09:56.750 ⇒ 00:09:58.409 Mustafa Raja: They use open telemet.
127 00:09:58.590 ⇒ 00:09:59.360 Mustafa Raja: telemetry.
128 00:09:59.360 ⇒ 00:09:59.990 Samuel Roberts: Yeah.
129 00:10:00.480 ⇒ 00:10:08.259 Samuel Roberts: Yeah, I… I didn’t have a ton of experience with that, beforehand, but, OpenTelemetry is, like.
130 00:10:09.200 ⇒ 00:10:12.230 Samuel Roberts: Pretty broad in what it can track.
131 00:10:12.560 ⇒ 00:10:12.940 Mustafa Raja: Hmm.
132 00:10:12.940 ⇒ 00:10:18.170 Samuel Roberts: And it’s got this concept of, spans… And…
133 00:10:18.540 ⇒ 00:10:20.089 Samuel Roberts: I forget what they call them now.
134 00:10:20.700 ⇒ 00:10:30.799 Samuel Roberts: But what I was getting was, like, I was getting all kinds of logs for, like, even just the Next.js stuff, so I was getting, like, requests and responses that weren’t LLM-related.
135 00:10:31.250 ⇒ 00:10:32.060 Mustafa Raja: Okay.
136 00:10:32.490 ⇒ 00:10:34.730 Samuel Roberts: And I was trying to fix that up, so.
137 00:10:35.620 ⇒ 00:10:38.369 Mustafa Raja: But I think once now… It’s compatible with Master, right.
138 00:10:38.940 ⇒ 00:10:41.900 Samuel Roberts: It looks like it is, based on seeing even this page here.
139 00:10:41.900 ⇒ 00:10:42.310 Mustafa Raja: Yeah.
140 00:10:42.310 ⇒ 00:10:44.169 Samuel Roberts: I’m seeing this guy here.
141 00:10:44.770 ⇒ 00:10:46.570 Samuel Roberts: So I’m hoping that that means it’ll…
142 00:10:47.900 ⇒ 00:10:52.349 Samuel Roberts: it’ll, play nicer than what I was seeing before, because I wasn’t even getting…
143 00:10:52.480 ⇒ 00:10:54.599 Samuel Roberts: Like, token counts and stuff.
144 00:10:54.860 ⇒ 00:10:55.650 Mustafa Raja: Hmm…
145 00:10:56.390 ⇒ 00:10:58.610 Samuel Roberts: And we want that eventually, so…
146 00:11:03.000 ⇒ 00:11:04.030 Samuel Roberts: Yeah…
147 00:11:15.640 ⇒ 00:11:17.340 Samuel Roberts: Excuse me.
148 00:11:18.920 ⇒ 00:11:19.670 Samuel Roberts: -
149 00:11:40.160 ⇒ 00:11:41.210 Samuel Roberts: The…
150 00:12:04.710 ⇒ 00:12:11.970 Samuel Roberts: I actually, while we’re waiting for a minute, might step away for a sec to grab some more coffee, but I’ll be back.
151 00:12:12.170 ⇒ 00:12:12.760 Mustafa Raja: Yep.
152 00:16:44.750 ⇒ 00:16:45.520 Uttam Kumaran: Hey, guys.
153 00:16:49.730 ⇒ 00:16:51.319 Uttam Kumaran: Hey, sorry for the delay.
154 00:16:54.870 ⇒ 00:16:59.330 Uttam Kumaran: Cool. Yeah, yesterday I just wanted to do, like, a brief…
155 00:16:59.760 ⇒ 00:17:07.180 Uttam Kumaran: print review, I blocked off some time… Bocked off some time…
156 00:17:07.520 ⇒ 00:17:16.389 Uttam Kumaran: In my schedule later today, to push some… Prs, so I’ll probably be…
157 00:17:16.650 ⇒ 00:17:20.859 Uttam Kumaran: being… be able to use the masterwork
158 00:17:21.319 ⇒ 00:17:26.599 Uttam Kumaran: and a couple other things, but maybe I just would love to just go through .
159 00:17:26.780 ⇒ 00:17:28.440 Samuel Roberts: And Rico, if you don’t…
160 00:17:28.530 ⇒ 00:17:33.599 Uttam Kumaran: Oh, you’re good. And Rico, if you don’t mind just keeping notes in the AI team?
161 00:17:34.440 ⇒ 00:17:35.700 Uttam Kumaran: print review.
162 00:17:36.060 ⇒ 00:17:38.709 Uttam Kumaran: That would be helpful.
163 00:17:39.100 ⇒ 00:17:42.650 Uttam Kumaran: So maybe I’m just gonna…
164 00:17:43.300 ⇒ 00:17:45.760 Uttam Kumaran: I have the linear board next to me, and then…
165 00:17:46.140 ⇒ 00:17:48.810 Uttam Kumaran: And just, like, click through a couple things.
166 00:17:50.750 ⇒ 00:17:52.540 Uttam Kumaran: Okay…
167 00:17:58.330 ⇒ 00:18:01.030 Uttam Kumaran: Okay, so I guess first thing…
168 00:18:01.540 ⇒ 00:18:07.770 Uttam Kumaran: we can just, I guess, just take a look at the, agentsMD work?
169 00:18:08.110 ⇒ 00:18:13.670 Uttam Kumaran: Data, or this is the Forge platform.
170 00:18:25.220 ⇒ 00:18:26.750 Uttam Kumaran: Nice. Solid.
171 00:18:26.980 ⇒ 00:18:33.399 Uttam Kumaran: Yeah, I think probably, one thing, Sam, I saw a tweet of a guy who actually had, like, a…
172 00:18:33.720 ⇒ 00:18:44.189 Uttam Kumaran: Sorry. He had a… he had a pre… Commit about actually updating like, the README and everything.
173 00:18:44.650 ⇒ 00:18:45.720 Samuel Roberts: Oh.
174 00:18:46.510 ⇒ 00:18:50.949 Uttam Kumaran: After every, like, new PR, basically, so that it stays up-to-date.
175 00:18:51.210 ⇒ 00:18:53.329 Samuel Roberts: That’s not a bad idea. Okay.
176 00:18:53.890 ⇒ 00:18:58.009 Uttam Kumaran: Maybe it’s something I’ll just dig up right now and see if I can send it to you.
177 00:18:58.350 ⇒ 00:19:04.029 Samuel Roberts: Yeah, that’d be great. But that is a… yeah, that’s one thing, like, READMEs and AgentMDs make me to… yeah, that’s smart.
178 00:19:06.410 ⇒ 00:19:08.420 Uttam Kumaran: Yeah, let’s see if I can…
179 00:19:12.980 ⇒ 00:19:14.899 Uttam Kumaran: See if I can find this…
180 00:19:39.050 ⇒ 00:19:42.419 Uttam Kumaran: Yeah, I’ll have to find it, but maybe we can create a ticket.
181 00:19:42.790 ⇒ 00:19:45.149 Uttam Kumaran: For that, which is just, like, how to keep…
182 00:19:47.140 ⇒ 00:19:50.630 Uttam Kumaran: The documentation, like, up-to-date on every…
183 00:19:51.510 ⇒ 00:19:52.080 Samuel Roberts: Yeah.
184 00:19:52.080 ⇒ 00:19:53.450 Uttam Kumaran: VR or commit.
185 00:19:54.970 ⇒ 00:19:58.889 Uttam Kumaran: So maybe that’s one thing I’ll throw in the backlog here.
186 00:19:59.320 ⇒ 00:20:02.859 Samuel Roberts: Yeah, I think PRs make more sense than commits, but definitely.
187 00:20:03.870 ⇒ 00:20:06.519 Samuel Roberts: Yeah, so, like, on PR…
188 00:20:06.830 ⇒ 00:20:10.930 Uttam Kumaran: Readmeagents.entity.
189 00:20:11.500 ⇒ 00:20:13.050 Uttam Kumaran: Docs, update.
190 00:20:14.680 ⇒ 00:20:15.400 Uttam Kumaran: Okay?
191 00:20:21.050 ⇒ 00:20:23.899 Uttam Kumaran: Okay, awesome. So yeah, I’m gonna… I’m gonna be…
192 00:20:24.390 ⇒ 00:20:31.530 Uttam Kumaran: Testing it out today, so I’ll see… Kinda like what the… How it works.
193 00:20:31.780 ⇒ 00:20:33.650 Samuel Roberts: Yeah, definitely, that’d be great.
194 00:20:33.920 ⇒ 00:20:44.349 Uttam Kumaran: Cool. And then… I guess, Casey, you worked on the export for the… Slack conversations, right?
195 00:20:44.780 ⇒ 00:20:49.649 Casie Aviles: Yes, I… I did create that, and I did do that, and then I passed it to…
196 00:20:49.830 ⇒ 00:20:52.090 Casie Aviles: AI for, like, a review.
197 00:20:53.030 ⇒ 00:20:58.060 Uttam Kumaran: Cool. Okay, great. Yeah, I wonder,
198 00:20:59.370 ⇒ 00:21:06.420 Uttam Kumaran: Yeah, I guess this is probably related to our, like, retrieve Slack messages…
199 00:21:06.600 ⇒ 00:21:08.850 Uttam Kumaran: API endpoint that we have to build, right?
200 00:21:10.800 ⇒ 00:21:12.959 Samuel Roberts: Yeah. What is the context of this one?
201 00:21:13.470 ⇒ 00:21:21.520 Uttam Kumaran: This one, we’re just… I just, we’re just doing, like, kind of, like, a review of some former contractors on just, like, what… what work did they do?
202 00:21:21.920 ⇒ 00:21:23.170 Samuel Roberts: Got it, okay, okay.
203 00:21:23.170 ⇒ 00:21:25.629 Uttam Kumaran: So I just needed an export of, like.
204 00:21:26.120 ⇒ 00:21:30.329 Uttam Kumaran: their Slack messages to see, because… Sometimes people, like.
205 00:21:31.540 ⇒ 00:21:36.480 Uttam Kumaran: We hire, like, as you guys know, like, we don’t… we were hiring pretty fast.
206 00:21:36.580 ⇒ 00:21:56.069 Uttam Kumaran: and now I think we’ve slowed down a bit, but, like, sometimes when people, like, for example, like, if people just start, and then they, like, it doesn’t work out, sometimes we go and, like, we just basically do an audit of, like, okay, what did they end up doing? Because usually people who are used to, like, consulting come in here, and then they just bill, like, 8 hours a day.
207 00:21:56.070 ⇒ 00:21:57.210 Samuel Roberts: And don’t do it.
208 00:21:57.670 ⇒ 00:21:58.640 Uttam Kumaran: And…
209 00:21:58.760 ⇒ 00:22:06.750 Uttam Kumaran: Like, yeah, at some people, it’s pretty egregious, and they’re stealing from us, so we are doing kind of a review,
210 00:22:07.250 ⇒ 00:22:08.050 Uttam Kumaran: Yeah.
211 00:22:08.560 ⇒ 00:22:10.810 Uttam Kumaran: Which sucks, like, I don’t want to do this type of work.
212 00:22:10.810 ⇒ 00:22:20.129 Samuel Roberts: Yeah, yeah, no, I was… yeah, because the endpoint for, like, the Client Hub stuff is all the embedding… all the embedded Slack messages, so I didn’t know if it was.
213 00:22:20.130 ⇒ 00:22:21.180 Uttam Kumaran: Oh, okay.
214 00:22:21.370 ⇒ 00:22:23.540 Samuel Roberts: So it’s a little different, is why I’m just pointing that out.
215 00:22:23.540 ⇒ 00:22:25.360 Uttam Kumaran: I see, I see, I see. Okay.
216 00:22:25.700 ⇒ 00:22:30.519 Uttam Kumaran: Yeah, I guess, all in all, this use case could have been nice to do
217 00:22:31.360 ⇒ 00:22:34.459 Uttam Kumaran: with that work, right? Somewhere in the UI, so…
218 00:22:35.120 ⇒ 00:22:37.180 Samuel Roberts: Yeah, and there’s definitely a way we could get that
219 00:22:37.960 ⇒ 00:22:43.639 Samuel Roberts: Hooked into, like, once these tools are kind of… like, there might be a nicer way to put, like, some monster tools together.
220 00:22:44.600 ⇒ 00:22:46.740 Samuel Roberts: Eventually, to make this.
221 00:22:47.940 ⇒ 00:22:49.200 Samuel Roberts: happen.
222 00:22:49.200 ⇒ 00:22:49.840 Uttam Kumaran: Yeah.
223 00:22:50.380 ⇒ 00:22:51.069 Samuel Roberts: But, yeah.
224 00:22:52.240 ⇒ 00:22:53.490 Uttam Kumaran: Okay, cool.
225 00:22:57.630 ⇒ 00:23:01.929 Uttam Kumaran: Great. Is the text the SQL…
226 00:23:02.100 ⇒ 00:23:06.960 Uttam Kumaran: Stuff, like, logged anywhere, or is it all just, like, on your local testing?
227 00:23:07.340 ⇒ 00:23:10.849 Samuel Roberts: It’s on my local… I was actually in the… that Eden.
228 00:23:10.850 ⇒ 00:23:11.360 Uttam Kumaran: Oh, great.
229 00:23:11.360 ⇒ 00:23:23.969 Samuel Roberts: Okay. Because I wanted it to be with the dbt stuff to start, at least. I haven’t pushed anything yet. So it’s all… and it’s a little… I’m just kind of working through some stuff there. I actually asked him a lot for some time, because I think…
230 00:23:24.370 ⇒ 00:23:29.759 Samuel Roberts: I… I don’t have a ton of contacts, and so the OOM also doesn’t have a ton of context, even looking at some of the dbt models of, like.
231 00:23:30.440 ⇒ 00:23:35.250 Samuel Roberts: more… like, it can read the dbt models and get some sense, but I think it’s still missing…
232 00:23:35.360 ⇒ 00:23:40.779 Samuel Roberts: Some, like, business context on top of that, that it would help it even to generate the…
233 00:23:40.900 ⇒ 00:23:44.800 Samuel Roberts: like, context file that we will use to generate the SQL, if that makes sense.
234 00:23:44.800 ⇒ 00:23:45.470 Uttam Kumaran: Yes.
235 00:23:45.820 ⇒ 00:23:48.210 Samuel Roberts: So I’m hoping to get some time with him later to, like.
236 00:23:48.330 ⇒ 00:23:53.280 Samuel Roberts: Get… the scaffolding is basically there for it, it just… it needs probably a…
237 00:23:53.580 ⇒ 00:23:55.419 Samuel Roberts: Someone with knowledge that is being…
238 00:23:55.420 ⇒ 00:23:58.429 Uttam Kumaran: I think you should book time with Awash and Demolade.
239 00:23:58.570 ⇒ 00:23:59.870 Samuel Roberts: Both of them? Okay. Yeah.
240 00:24:00.190 ⇒ 00:24:02.470 Uttam Kumaran: Do you want to do that? Like, do you want me to.
241 00:24:02.470 ⇒ 00:24:03.000 Samuel Roberts: Yeah, definitely.
242 00:24:03.000 ⇒ 00:24:03.770 Uttam Kumaran: that, or…
243 00:24:04.290 ⇒ 00:24:08.499 Samuel Roberts: Yeah, I just sent him a message this morning, I don’t think I even heard back yet, but I will see if I can grab time with both of them.
244 00:24:08.500 ⇒ 00:24:13.260 Uttam Kumaran: Yeah, if you got… if you three can meet today, that’d be so helpful, because I’m sort of slammed, but that would be.
245 00:24:13.260 ⇒ 00:24:14.600 Samuel Roberts: Yeah, definitely. Okay.
246 00:24:15.500 ⇒ 00:24:16.450 Uttam Kumaran: Amazing.
247 00:24:18.580 ⇒ 00:24:20.090 Uttam Kumaran: Okay.
248 00:24:21.360 ⇒ 00:24:25.079 Uttam Kumaran: Hold on one second, my dog just, like, ran out of the house. One second, let me just go quick.
249 00:24:25.080 ⇒ 00:24:26.090 Samuel Roberts: Oh, no.
250 00:27:01.360 ⇒ 00:27:08.810 Uttam Kumaran: Okay, next thing was… the…
251 00:27:09.370 ⇒ 00:27:12.839 Uttam Kumaran: Okay, so we talked about the Master Migration Plan.
252 00:27:13.030 ⇒ 00:27:23.420 Uttam Kumaran: Texas SQL, so that one, I think, is basically… Gonna move to… escalation and review,
253 00:27:23.750 ⇒ 00:27:28.279 Uttam Kumaran: Rico, can you send… can you send a thread in engineering and see if,
254 00:27:29.210 ⇒ 00:27:33.140 Uttam Kumaran: You can help coordinate with Stam on a time for everybody to meet.
255 00:27:35.490 ⇒ 00:27:42.370 Uttam Kumaran: Okay, and then… yeah, a couple things,
256 00:27:43.150 ⇒ 00:27:47.449 Uttam Kumaran: So, let me… let me sort of pitch you guys a little bit on, like.
257 00:27:48.030 ⇒ 00:27:51.070 Uttam Kumaran: what I’m gonna try to work on today, so…
258 00:27:51.550 ⇒ 00:28:02.009 Uttam Kumaran: kind of, like, how I’m thinking about it is really aligned to… are… Docs here,
259 00:28:03.440 ⇒ 00:28:08.290 Uttam Kumaran: So, I’m sort of… now that we have a pretty fixed structure for…
260 00:28:08.660 ⇒ 00:28:12.960 Uttam Kumaran: project management, and, like, kind of some of the tasks, and I’ll update this.
261 00:28:13.360 ⇒ 00:28:21.790 Uttam Kumaran: I almost want to create… at least, like, and this is where, like, I’m gonna bias towards… like…
262 00:28:22.050 ⇒ 00:28:23.440 Uttam Kumaran: Too many.
263 00:28:24.290 ⇒ 00:28:29.310 Uttam Kumaran: UIs… But they’re very use case specific, and then…
264 00:28:30.120 ⇒ 00:28:35.139 Uttam Kumaran: I… I think what we’re lacking at the moment is, like, the fundamental…
265 00:28:35.650 ⇒ 00:28:38.809 Uttam Kumaran: solutions for some of these problems, and then I can let
266 00:28:39.080 ⇒ 00:28:45.399 Uttam Kumaran: you guys decide on how to consolidate, but for example, I would like there to be something that helps support the
267 00:28:45.880 ⇒ 00:28:50.059 Uttam Kumaran: The project management team during and after stand-ups.
268 00:28:50.230 ⇒ 00:28:55.050 Samuel Roberts: Right? And there’s very clear things that we do during stand-ups now, which is, like, we review tickets.
269 00:28:55.050 ⇒ 00:28:58.270 Uttam Kumaran: Review new things that came up. Those are all possible.
270 00:28:58.610 ⇒ 00:29:05.129 Uttam Kumaran: I want us to support grooming, and there’s a set of activities that the PMs do during grooming, like.
271 00:29:05.660 ⇒ 00:29:11.730 Uttam Kumaran: Review backlog… like, help create new tickets, so I kind of want to create almost, like.
272 00:29:12.130 ⇒ 00:29:20.790 Uttam Kumaran: an assistant for each step of the PM process, and that will be a collection of, like, nifty UIs.
273 00:29:21.070 ⇒ 00:29:27.319 Uttam Kumaran: some display elements, and then calling some APIs from, like, linear and things like that.
274 00:29:27.690 ⇒ 00:29:31.590 Uttam Kumaran: And then I want something to… so basically the kind of the four elements are, like.
275 00:29:32.160 ⇒ 00:29:38.079 Uttam Kumaran: There’s a couple of them. There’s planning, stand-ups, grooming, print review.
276 00:29:38.290 ⇒ 00:29:40.939 Uttam Kumaran: And then there’s also client meeting.
277 00:29:41.200 ⇒ 00:29:47.590 Uttam Kumaran: So, like, before a client meeting, we want to have a helpful, like, set of AI tools that helps, like.
278 00:29:47.680 ⇒ 00:30:05.850 Uttam Kumaran: the PM and the team prepare for, like, what we’re gonna talk about. To give you a sense of, like, why I think it’s… one, we do have… instead of just supporting, like, a mix of tools everywhere for everybody, I think my new hypothesis is, like, support the parts of the delivery process that we know are pretty fixed.
279 00:30:06.170 ⇒ 00:30:15.270 Uttam Kumaran: and just build everything that’s needed to do that. And in fact, I’ve been doing a lot of this stuff this week in just my local ChatGPT,
280 00:30:15.610 ⇒ 00:30:21.599 Uttam Kumaran: Which is, like, copying and pasting a bunch of things, and putting a pre-selected prompt in.
281 00:30:21.830 ⇒ 00:30:26.690 Uttam Kumaran: Which is, like… yeah, so I feel like I have the building blocks.
282 00:30:27.050 ⇒ 00:30:33.060 Uttam Kumaran: to support each of those core activities, and, like, that’s the sort of stuff I’m gonna try to work on today.
283 00:30:33.370 ⇒ 00:30:40.589 Uttam Kumaran: And so… and then, similarly on the sales side, it’s very similar, like, we have…
284 00:30:40.730 ⇒ 00:30:50.270 Uttam Kumaran: Sales follow-ups, the sales meeting, like, SOW preparation, all of that, and I want to sort of almost adhere
285 00:30:50.650 ⇒ 00:30:51.710 Uttam Kumaran: the
286 00:30:52.920 ⇒ 00:31:02.019 Uttam Kumaran: pages towards, like, an activity that’s happening. And because there are pretty distinct activities that happen across sales and PM,
287 00:31:02.230 ⇒ 00:31:12.990 Uttam Kumaran: that I think, we can start to just build, like, a page for that activity. Every time you’re going to make an SOW, this is the SOW building interface.
288 00:31:13.150 ⇒ 00:31:16.250 Uttam Kumaran: Or, like, the assistant interface for SOW Building.
289 00:31:17.260 ⇒ 00:31:18.810 Samuel Roberts: Trusting tasks, like.
290 00:31:18.810 ⇒ 00:31:19.280 Uttam Kumaran: Yes.
291 00:31:19.280 ⇒ 00:31:20.949 Samuel Roberts: Okay, okay, cool.
292 00:31:20.950 ⇒ 00:31:21.730 Uttam Kumaran: Pretty distinct.
293 00:31:21.730 ⇒ 00:31:23.880 Samuel Roberts: Between pages as much as… yeah.
294 00:31:24.070 ⇒ 00:31:29.489 Uttam Kumaran: Yeah, pretty distinct tasks that… Have a set outcome.
295 00:31:30.070 ⇒ 00:31:31.240 Samuel Roberts: Yeah, okay, go ahead.
296 00:31:31.240 ⇒ 00:31:36.730 Uttam Kumaran: That require… may require… Like, tons of different data, basically.
297 00:31:36.730 ⇒ 00:31:41.339 Samuel Roberts: That’s, yeah, exactly, yeah, but it’s still, like, doing one thing, but pulling stuff from one place and getting an output.
298 00:31:42.980 ⇒ 00:31:48.129 Uttam Kumaran: That’s where I don’t know, like, this is where, for some of the PM stuff, it may involve linear…
299 00:31:48.280 ⇒ 00:31:49.810 Uttam Kumaran: Clockify…
300 00:31:50.210 ⇒ 00:31:51.290 Samuel Roberts: Black…
301 00:31:51.830 ⇒ 00:31:52.590 Uttam Kumaran: up.
302 00:31:52.900 ⇒ 00:31:54.570 Uttam Kumaran: pre-filled prompt…
303 00:31:55.290 ⇒ 00:32:02.100 Uttam Kumaran: documents… so what I’m… what you’re… what we’re probably gonna see my asks for is gonna be to…
304 00:32:02.420 ⇒ 00:32:05.109 Uttam Kumaran: develop those, like, APIs.
305 00:32:06.020 ⇒ 00:32:09.290 Uttam Kumaran: And then probably the next thing, Sam, to think about is, like.
306 00:32:09.620 ⇒ 00:32:15.330 Uttam Kumaran: How do you do a gentic search over, like, large… data sets.
307 00:32:15.480 ⇒ 00:32:19.980 Uttam Kumaran: like… Where you can’t push all the context in.
308 00:32:20.180 ⇒ 00:32:21.530 Samuel Roberts: Yeah.
309 00:32:22.060 ⇒ 00:32:40.089 Uttam Kumaran: You know, so, like, embedding is one thing. I think that’s gonna be more of my asks, which is kind of aligned with what we said, which is, like, we build a rich set of APIs, document them, and they can get utilized a multitude of ways. But I think what we’ll see here is most of the challenge, now that we have Mastra, is
310 00:32:40.170 ⇒ 00:32:44.679 Uttam Kumaran: It’s gonna be more on the infrastructure and the data engineering side.
311 00:32:45.030 ⇒ 00:32:45.470 Samuel Roberts: Yeah, the.
312 00:32:45.470 ⇒ 00:32:48.629 Uttam Kumaran: To make sure that those APIs work, the search APIs work.
313 00:32:48.790 ⇒ 00:32:52.920 Uttam Kumaran: for example, for the marketing team, if I want to build a, like.
314 00:32:53.110 ⇒ 00:32:56.570 Uttam Kumaran: one of the things they do is they have a GPD that,
315 00:32:56.800 ⇒ 00:32:59.710 Uttam Kumaran: They have a GPT that generates
316 00:33:00.300 ⇒ 00:33:12.720 Uttam Kumaran: like, we bid on Upwork and other marketplace jobs. They have a GPT that helps them bid on those, but what does it do? It pulls from a prompt, and it pulls from around 30 different documents to select the right marketing assets to link.
317 00:33:13.320 ⇒ 00:33:13.950 Samuel Roberts: Right.
318 00:33:13.950 ⇒ 00:33:20.849 Uttam Kumaran: But my question there is the prompt, like, that’s totally possible, but, like, how can we support pulling…
319 00:33:21.140 ⇒ 00:33:26.149 Uttam Kumaran: from the marketing assets document list. Okay, so then my ask is gonna be.
320 00:33:26.360 ⇒ 00:33:28.920 Uttam Kumaran: We… someone needs to probably build a…
321 00:33:29.300 ⇒ 00:33:36.009 Uttam Kumaran: get relevant, or whatever, it’s like a search function for the marketing assets, that is better
322 00:33:36.130 ⇒ 00:33:45.549 Uttam Kumaran: that performs better than the one in ChatGPT, you know? Right. And the ChatGPT one is just stuffing everything in context. This should have some better ability to do this search.
323 00:33:45.890 ⇒ 00:33:56.670 Uttam Kumaran: And then, nicely, what’s gonna happen is that GetMaterials endpoint can be used in any sort of sales context, right? So, that’s what I’m hoping for.
324 00:33:59.360 ⇒ 00:34:00.830 Samuel Roberts: Yes. Yeah, I know you.
325 00:34:00.830 ⇒ 00:34:01.600 Uttam Kumaran: You’re finding them in Pico.
326 00:34:01.600 ⇒ 00:34:06.549 Samuel Roberts: And they very well may be endpoints, but they also just might be tools in Monstra that we can piece together.
327 00:34:06.740 ⇒ 00:34:09.650 Uttam Kumaran: Exactly, yeah, so I’m talking to them endpoint… Yeah, yeah, yeah.
328 00:34:09.659 ⇒ 00:34:12.999 Samuel Roberts: And they very well may be endpoints if we need, like, external things to call them.
329 00:34:12.999 ⇒ 00:34:18.729 Uttam Kumaran: But the nice thing about Monster being in the platform is we don’t need to expose them, necessarily, and have extra… Okay.
330 00:34:18.729 ⇒ 00:34:27.799 Samuel Roberts: Okay. But I understand, I’m not… I get exactly what you’re saying. I’m thinking… I’m thinking of them as, like, tools, agents, and workflows, which are the kind of master… Yes.
331 00:34:28.109 ⇒ 00:34:39.889 Samuel Roberts: And, like, those all can be endpoints, if, for example, like, an NHN thing needs it, or, like, you know, Linear wants to hit something, or Slack needs to hit something, but if it’s a tool on the forge, then it’s all kind of…
332 00:34:40.009 ⇒ 00:34:48.499 Samuel Roberts: it’s still an endpoint, because it’s technically a server and everything, but all this stuff happens, and yeah. I’m not being nitpicky, I just want to let you know, like, I understand what you’re saying here.
333 00:34:49.940 ⇒ 00:34:51.619 Uttam Kumaran: Yeah. No, no, you’re right.
334 00:34:52.940 ⇒ 00:34:57.890 Uttam Kumaran: Okay, great, so I think that’ll help give some more…
335 00:34:58.290 ⇒ 00:35:01.490 Uttam Kumaran: Structure to the asks versus developing
336 00:35:01.660 ⇒ 00:35:04.890 Uttam Kumaran: Like, infrastructure with no clear use case.
337 00:35:05.200 ⇒ 00:35:05.700 Samuel Roberts: Yeah.
338 00:35:05.700 ⇒ 00:35:09.039 Uttam Kumaran: So I’m gonna go… I’m gonna try to go the distance on one or two of those.
339 00:35:09.340 ⇒ 00:35:11.449 Uttam Kumaran: And the nice thing is, like, I think I can…
340 00:35:11.590 ⇒ 00:35:15.450 Uttam Kumaran: I’ll do some basic UI, and then we have the Mostra.
341 00:35:15.560 ⇒ 00:35:20.059 Uttam Kumaran: And then we have the Agent 70, so I think I have enough, and then, so I’ll kind of share how it goes.
342 00:35:20.260 ⇒ 00:35:21.350 Samuel Roberts: Okay, yes.
343 00:35:23.570 ⇒ 00:35:26.249 Uttam Kumaran: You know, and then we can start to layer things on.
344 00:35:27.010 ⇒ 00:35:28.250 Samuel Roberts: Yeah, I would say.
345 00:35:28.250 ⇒ 00:35:32.150 Uttam Kumaran: I did a lot, I did a lot of ChatGPT work this week.
346 00:35:32.400 ⇒ 00:35:35.120 Uttam Kumaran: For a ton of different activities, like…
347 00:35:35.120 ⇒ 00:35:36.380 Samuel Roberts: Okay.
348 00:35:36.770 ⇒ 00:35:38.210 Uttam Kumaran: For example, like.
349 00:35:38.740 ⇒ 00:35:45.019 Uttam Kumaran: This is for a specific deal, where I just created a custom GPT, uploaded all the documents related to the deal.
350 00:35:45.200 ⇒ 00:35:51.559 Uttam Kumaran: This is something that, like… Totally could have been better, given, like, all of our other information.
351 00:35:51.770 ⇒ 00:35:58.619 Uttam Kumaran: I was able to create different templates in here, like, I had something help me with creating
352 00:35:59.130 ⇒ 00:36:04.170 Uttam Kumaran: data documents, and so I think what I’m… what I’m gonna understand is how much of this
353 00:36:04.870 ⇒ 00:36:16.209 Uttam Kumaran: the problem with me doing it in here is that I… it’s hard to get other people to also come see it in here. And so, one of the… that’s the kind of the thing I’m trying to solve for, which is, like.
354 00:36:16.760 ⇒ 00:36:22.310 Uttam Kumaran: some of the activities I was doing were kind of distinct, like, we have a deal, I need to…
355 00:36:22.810 ⇒ 00:36:27.780 Uttam Kumaran: I need to produce an output, like, an SOW or something, And…
356 00:36:28.280 ⇒ 00:36:42.820 Uttam Kumaran: I have the format of the SOW, but I wanted to consider, like, a bunch of other documents, or the ways about Brainforge, and, you know, I’m doing that just in a simple prompt here versus, like, anything more sophisticated, so…
357 00:36:43.650 ⇒ 00:36:52.959 Samuel Roberts: is it worth spending some time listing all that stuff out before you start, like, digging into specific things, you think? Like, now that you’ve done a bunch of stuff this week.
358 00:36:53.270 ⇒ 00:37:00.200 Uttam Kumaran: Yeah, so what I’m gonna start… Yeah, so this week, I’m gonna start… I’m gonna start everything with, like, a PRD for one of these.
359 00:37:00.200 ⇒ 00:37:00.760 Samuel Roberts: Okay, cool.
360 00:37:00.760 ⇒ 00:37:08.040 Uttam Kumaran: And I don’t think we have, like, a great ERD format, but I’m… I may do it in…
361 00:37:08.840 ⇒ 00:37:15.080 Uttam Kumaran: I may do it in Figma, or I may do it in Notion, but yeah, everything is gonna start with, like, product requirements first.
362 00:37:15.080 ⇒ 00:37:15.670 Samuel Roberts: Okay.
363 00:37:16.170 ⇒ 00:37:18.630 Uttam Kumaran: And then…
364 00:37:20.090 ⇒ 00:37:26.770 Uttam Kumaran: move, and then from there, that way, that’s what I’m gonna hand… that’s what I’m going to hand to…
365 00:37:26.900 ⇒ 00:37:27.950 Uttam Kumaran: Codex.
366 00:37:28.360 ⇒ 00:37:28.840 Samuel Roberts: Yeah.
367 00:37:28.840 ⇒ 00:37:30.119 Uttam Kumaran: And then that way.
368 00:37:30.530 ⇒ 00:37:43.790 Uttam Kumaran: you can go review my seed… that’s the seed file, right? So you can go review that to understand, like, what the heck the goal is. Because there will be a MVP… there’ll be a proof of concept MVP V1, V2, for every single PRD.
369 00:37:43.790 ⇒ 00:37:44.300 Samuel Roberts: Right.
370 00:37:44.400 ⇒ 00:37:48.310 Uttam Kumaran: Yeah, so… I will… I’m gonna start everything from there.
371 00:37:48.770 ⇒ 00:37:53.910 Samuel Roberts: Okay, cool. Yeah, my thought is… is, one, it’s good for that, and two, as you start
372 00:37:54.140 ⇒ 00:38:10.949 Samuel Roberts: like, listing those things out, and you might realize, like, okay, this tool is one that can be used both places, and so, like, I’m a little worried Codex might go a little crazy creating tools and master agents, so I’m just trying to think, like, if we can point it even to, like, once you’ve run one of them.
373 00:38:10.950 ⇒ 00:38:15.350 Samuel Roberts: you might update that AgentsMD to have the new tool
374 00:38:15.680 ⇒ 00:38:18.440 Samuel Roberts: in Mastra that it can make use of, or something like that.
375 00:38:18.440 ⇒ 00:38:19.100 Uttam Kumaran: I see.
376 00:38:19.100 ⇒ 00:38:31.290 Samuel Roberts: Kind of like… I don’t know if they have water flow, but that has other connotations, but the idea being, like, it doesn’t try to, like, reinvent the wheel when you run it on the next one, even if you already have some infrastructure in place for a similar kind of thing.
377 00:38:31.940 ⇒ 00:38:32.920 Uttam Kumaran: Okay, okay.
378 00:38:33.210 ⇒ 00:38:39.660 Samuel Roberts: But that might be a little more me just trying to future-proof, which it might be not worth it, but,
379 00:38:40.390 ⇒ 00:38:47.249 Samuel Roberts: I’m just… yeah, that’s how my brain works, so I don’t know. Okay. But yeah, give it a try and see, I mean, there’s no… it’s not gonna hurt, you know, so…
380 00:38:47.570 ⇒ 00:38:51.000 Samuel Roberts: Worst case, it makes two of the same tools, and we just end up, you know…
381 00:38:51.560 ⇒ 00:38:57.079 Uttam Kumaran: Yeah, that’s the thing, like, I’m gonna bias towards having something that, like, works, and then…
382 00:38:57.250 ⇒ 00:39:01.380 Samuel Roberts: Because that is the limiting factor, is, like, the vision of, like, what it is.
383 00:39:01.460 ⇒ 00:39:06.609 Uttam Kumaran: And then I’ll be like, hey, it’s working… this whole setup is, like, roughly working for me.
384 00:39:06.790 ⇒ 00:39:13.939 Uttam Kumaran: Can you check out? And then what you could… ideally, what you should do is then see, like, oh, codex went overboard here, we should put it in the agent’s MD file.
385 00:39:13.940 ⇒ 00:39:19.730 Samuel Roberts: Exactly, that’s what… yeah, I think I’m just thinking that step forward, and I honestly, it shouldn’t really matter, I can probably just take care of that, so you’re right.
386 00:39:20.140 ⇒ 00:39:25.370 Uttam Kumaran: Yeah, and so it’ll be… it’ll be clunky to start, and then… Naturally, like, we’ll move from…
387 00:39:26.310 ⇒ 00:39:31.959 Uttam Kumaran: like, basically how I’m gonna do this is I’m gonna create a pretty distinct PRD format, I’m gonna talk to…
388 00:39:32.330 ⇒ 00:39:41.789 Uttam Kumaran: ChatGPT, probably for, like, 3-4 minutes at length, about, like, the problems I’m facing, the user store… like, I’ll just use Whisper.
389 00:39:41.960 ⇒ 00:39:49.039 Uttam Kumaran: and the impact, and examples of where I did it before. I’ll have it write the PRD, I’ll edit it.
390 00:39:49.280 ⇒ 00:39:55.089 Uttam Kumaran: kind of make sure that it has all the core functionality, and then I’ll hand it to Codex to develop.
391 00:39:55.370 ⇒ 00:39:57.429 Uttam Kumaran: the… that’s the thing, though, I…
392 00:39:57.590 ⇒ 00:40:02.950 Uttam Kumaran: at the time I hand it to Codex, I don’t know what it knows from the codebase.
393 00:40:03.190 ⇒ 00:40:06.129 Uttam Kumaran: So, we’ll see. I don’t think… I’m not…
394 00:40:06.300 ⇒ 00:40:15.509 Uttam Kumaran: we’ll see how close it gets to, like, one-shotting a simple proof of concept or something. Like, I don’t think it’s gonna get… this is where, like, even in the PRD, I’m gonna give it design.
395 00:40:16.300 ⇒ 00:40:16.890 Samuel Roberts: I will use.
396 00:40:16.890 ⇒ 00:40:18.840 Uttam Kumaran: use, I’ll either…
397 00:40:19.170 ⇒ 00:40:27.480 Uttam Kumaran: describe the design, do a simple wireframe in Figma using, like, the Figma MCP, or I will use magic patterns.
398 00:40:27.930 ⇒ 00:40:34.830 Uttam Kumaran: So it’ll be a serious PRD, like, it’ll be, like, a pretty legit, like, normal product PRD. That way.
399 00:40:35.030 ⇒ 00:40:37.840 Uttam Kumaran: You can always, like, trace back.
400 00:40:37.950 ⇒ 00:40:44.380 Uttam Kumaran: Because without the traceability, like, if you trace it back to me being like, build this, there’s no traceability, there’s no reproducibility, so…
401 00:40:44.380 ⇒ 00:40:44.830 Samuel Roberts: Definitely.
402 00:40:44.830 ⇒ 00:40:45.530 Uttam Kumaran: Yeah.
403 00:40:46.350 ⇒ 00:40:51.429 Uttam Kumaran: And then the last thing I wanted to talk about was, like, we paid for LankFuse, and so, like…
404 00:40:51.430 ⇒ 00:40:51.950 Samuel Roberts: Yeah, weird.
405 00:40:51.950 ⇒ 00:40:58.080 Uttam Kumaran: I think what, like, that’s also something I think next week we can talk about, like, LankView setup, probably.
406 00:40:58.280 ⇒ 00:40:59.110 Uttam Kumaran: Right?
407 00:40:59.900 ⇒ 00:41:04.219 Samuel Roberts: Yeah, yeah, we were talking about it a little bit before you hopped on, but I saw the invite and stuff.
408 00:41:05.400 ⇒ 00:41:06.830 Mustafa Raja: Yeah, nice.
409 00:41:06.830 ⇒ 00:41:07.950 Samuel Roberts: Go ahead, go ahead, Mustafa.
410 00:41:07.950 ⇒ 00:41:15.730 Mustafa Raja: Yeah, Sam sent a link in chat. It does integrate with Master, so we can have observability of master over.
411 00:41:15.870 ⇒ 00:41:17.010 Mustafa Raja: In,
412 00:41:17.010 ⇒ 00:41:17.560 Uttam Kumaran: Nice.
413 00:41:17.560 ⇒ 00:41:22.169 Mustafa Raja: instance, and it’s a pretty quick setup also. Yeah.
414 00:41:22.380 ⇒ 00:41:25.449 Mustafa Raja: Like, a few lines only, 2, 3, yeah.
415 00:41:25.930 ⇒ 00:41:26.950 Mustafa Raja: Pretty simple.
416 00:41:27.870 ⇒ 00:41:28.509 Samuel Roberts: That’d be pretty good.
417 00:41:28.510 ⇒ 00:41:33.769 Uttam Kumaran: Yeah, I have a friend that’s using Langfuse in production at a company.
418 00:41:33.970 ⇒ 00:41:38.270 Uttam Kumaran: like, a friend company of ours. I was gonna call him just to, like, get…
419 00:41:38.960 ⇒ 00:41:42.310 Uttam Kumaran: You may have worked with Casey Patrick, remember Patrick?
420 00:41:42.310 ⇒ 00:41:43.270 Casie Aviles: Yeah, yeah.
421 00:41:43.570 ⇒ 00:41:50.410 Uttam Kumaran: Yeah, he’s… they’re doing… they’re using Langfuse for a bunch of stuff at… at… at Wild, and I may call him next week.
422 00:41:51.440 ⇒ 00:41:51.900 Mustafa Raja: Hmm.
423 00:41:51.900 ⇒ 00:41:59.349 Uttam Kumaran: Maybe, Rico, can we… can you send a note in AI Team, and then I’ll try to find a time, and I’ll have them hop on and just give us, like, a…
424 00:41:59.740 ⇒ 00:42:01.720 Uttam Kumaran: Tips and tricks for a length fuse.
425 00:42:04.680 ⇒ 00:42:08.439 Casie Aviles: Yeah, we considered long fees back then, as well.
426 00:42:08.440 ⇒ 00:42:09.250 Uttam Kumaran: Yeah.
427 00:42:09.460 ⇒ 00:42:14.840 Casie Aviles: we didn’t… it didn’t have, like, N8N integrations, and we were heavy into N8N.
428 00:42:14.840 ⇒ 00:42:15.940 Uttam Kumaran: Yeah, yeah.
429 00:42:15.940 ⇒ 00:42:16.710 Casie Aviles: for.
430 00:42:16.710 ⇒ 00:42:19.330 Samuel Roberts: Yeah, I think frantic, it’s still not the most ideal
431 00:42:19.660 ⇒ 00:42:28.100 Samuel Roberts: certain things. The prompt injection, it did pretty well, because it had a… it had a node, but for evaluations and tracing and stuff, I don’t think it would have worked.
432 00:42:28.100 ⇒ 00:42:29.149 Uttam Kumaran: Yeah. But, yeah.
433 00:42:29.600 ⇒ 00:42:40.839 Samuel Roberts: But for Mastra, it looks like it plugs in. I was trying to get it set up with Copilot Kit, but it was a little weird, but Copilot Kit with Mastra, everything will run through the Mastra agents, so it should be good.
434 00:42:41.140 ⇒ 00:42:42.080 Uttam Kumaran: Sick, okay.
435 00:42:43.810 ⇒ 00:42:44.930 Uttam Kumaran: Okay, sick.
436 00:42:45.400 ⇒ 00:42:49.010 Uttam Kumaran: Let’s pray for no more, framework changes.
437 00:42:51.270 ⇒ 00:42:51.969 Samuel Roberts: Yeah, I mean.
438 00:42:51.970 ⇒ 00:42:52.340 Uttam Kumaran: Honestly.
439 00:42:52.340 ⇒ 00:42:52.970 Samuel Roberts: You guys are.
440 00:42:52.970 ⇒ 00:42:55.890 Uttam Kumaran: You guys are, like, laughing, crying.
441 00:42:55.890 ⇒ 00:43:06.159 Samuel Roberts: Yeah, no, honestly, I mean, here’s the thing, and I had… I was talking with Hannah yesterday about, like, NHN and stuff, like, it’s great for certain things, and it’s also, like… it’s great for a lot of things, but it’s also…
442 00:43:07.620 ⇒ 00:43:17.329 Samuel Roberts: not always the most ideal tool, you know? Nothing is the… nothing is… this is just, you know, my kind of resolution was, like, you gotta find the right tool for the job, and that’s very dependent on lots of things, including, like.
443 00:43:17.330 ⇒ 00:43:25.259 Uttam Kumaran: No, like, we are… we are engineers, and, like, so we have a higher level alpha, which is, like, actually writing code. Yeah. Like, N8N…
444 00:43:25.420 ⇒ 00:43:28.499 Uttam Kumaran: the reason for it is…
445 00:43:28.640 ⇒ 00:43:41.940 Uttam Kumaran: because they have a fundamental belief that not everyone has to write code, and they want to support that. Well, we are, like, we just have a higher, like, we can do that, and then once we find that there are higher level abstractions, code is…
446 00:43:42.090 ⇒ 00:43:47.080 Samuel Roberts: the higher density language to do these things in, versus visual, like…
447 00:43:47.140 ⇒ 00:43:50.150 Uttam Kumaran: That’s… That’s all it is. So for… for…
448 00:43:50.440 ⇒ 00:44:04.959 Uttam Kumaran: the general, like, the addressable market of people that want… that can use low-code versus code is much bigger. And so you’re gonna hear them more, you’re gonna hear that N8N is the best thing ever, because there’s more of those people.
449 00:44:05.070 ⇒ 00:44:13.320 Uttam Kumaran: But I don’t, like, as usual, I don’t really care what other people say. It’s like, what is working for us? And we may use both.
450 00:44:13.680 ⇒ 00:44:14.589 Samuel Roberts: Yeah, I forget what…
451 00:44:14.590 ⇒ 00:44:17.619 Uttam Kumaran: Ryan is doing a bunch of nice HubSpot automation.
452 00:44:17.800 ⇒ 00:44:26.969 Uttam Kumaran: in NMN, and, like, you should keep doing that. But there are some fundamental things that are running the backbone of our company that are in, like, a visual diagramming layer. We should move that shit out, you know?
453 00:44:26.990 ⇒ 00:44:39.350 Samuel Roberts: 100%. Yep, that’s exactly what I was thinking. Like, it’s great for prototyping, it’s great for… it’s… honestly, it’s nice, it’s not even running on our infrastructure, we don’t have to worry about, you know, Heroku stuff for it. Like, there’s a lot of other benefits, but it also has limitations because of that, and I think we’re.
454 00:44:39.350 ⇒ 00:44:39.730 Uttam Kumaran: Yeah.
455 00:44:39.730 ⇒ 00:44:41.519 Samuel Roberts: Now, so, yeah, this is great.
456 00:44:42.430 ⇒ 00:44:43.470 Samuel Roberts: Okay.
457 00:44:44.040 ⇒ 00:44:44.690 Uttam Kumaran: Cool.
458 00:44:44.690 ⇒ 00:44:49.980 Samuel Roberts: Alright, so we have the next meeting, should be kind of dense, if you guys can join that, that’d be really great, and then… Cool.
459 00:44:50.330 ⇒ 00:44:52.640 Uttam Kumaran: Yeah, okay, perfect. Thank you, guys.
460 00:44:52.640 ⇒ 00:44:53.340 Samuel Roberts: See you there.
461 00:44:53.690 ⇒ 00:44:54.770 Uttam Kumaran: Okay, bye.
462 00:44:55.140 ⇒ 00:44:55.700 Mustafa Raja: Right.