Meeting Title: Eden - Brainforge: AI Command Center Weekly Check-in Date: 2026-05-08 Meeting participants: Pranav, Daniel
WEBVTT
1 00:01:49.400 ⇒ 00:01:50.290 Did you nope.
2 00:02:11.500 ⇒ 00:02:12.140 Daniel: Enough!
3 00:02:14.470 ⇒ 00:02:16.219 Pranav: Hey, Daddy, how’s it going?
4 00:02:16.360 ⇒ 00:02:17.550 Daniel: What’s going on, man?
5 00:02:17.950 ⇒ 00:02:20.709 Pranav: Nothing much, nothing much, nice to see ya.
6 00:02:20.980 ⇒ 00:02:24.609 Daniel: Good to see you, too! You guys have been crankin’.
7 00:02:25.170 ⇒ 00:02:32.089 Pranav: We have, we have. We’ve been… we’ve been working, we’ve been working. As you have been, too, right? You’ve been… you’ve been kind of spread thin, it sounds like.
8 00:02:32.650 ⇒ 00:02:37.410 Daniel: Oh, yeah, I mean, you know, as always, but this one, this one…
9 00:02:37.570 ⇒ 00:02:44.509 Daniel: This one was pretty fun. I have to say, you sent me the pulse email a little bit, like, just a little while ago.
10 00:02:44.510 ⇒ 00:02:45.100 Pranav: Yeah.
11 00:02:45.100 ⇒ 00:02:46.529 Daniel: That’s pretty impressive.
12 00:02:47.080 ⇒ 00:02:49.109 Pranav: You like it? Okay, cool. Cool, cool, cool.
13 00:02:49.110 ⇒ 00:02:51.120 Daniel: Impressive. Yeah.
14 00:02:51.720 ⇒ 00:02:55.150 Daniel: I read it, and it’s like, oh man, that’s like,
15 00:02:56.690 ⇒ 00:03:02.100 Daniel: what… the context that was given on this, like, how’s it picking up? Like, it’s doing some cool stuff.
16 00:03:03.140 ⇒ 00:03:15.740 Pranav: It is. So, I’m happy to kind of get into, like, how we were able to, like, drive this as well, if you kind of want to nerd out a little bit. But, like, yeah, let’s… let’s save some time for that at some point.
17 00:03:16.070 ⇒ 00:03:20.309 Pranav: Let’s just hop into it, let’s see if Adam joins in a little bit. I don’t think Robert’s going to.
18 00:03:20.310 ⇒ 00:03:27.559 Daniel: In fact, it was able to pull out and position, like, Danny’s upset about this. I was like, wow.
19 00:03:29.430 ⇒ 00:03:30.180 Pranav: Yeah.
20 00:03:31.480 ⇒ 00:03:34.160 Pranav: No, no, it’s, it’s interesting, and it’s,
21 00:03:35.000 ⇒ 00:03:40.850 Pranav: What we did was, basically, we did, like, semantic searching on the meeting notes.
22 00:03:40.970 ⇒ 00:03:43.370 Pranav: Specifically. And so…
23 00:03:43.820 ⇒ 00:04:00.839 Pranav: That… that is huge, you know, we’ve done that on other projects for, like, other clients before, and it really is, like, a huge driving factor, specifically for, like, theme analysis, and also just detecting, like, okay, what are the… the overarching, like, kind of, like, summarizations of certain calls?
24 00:04:00.870 ⇒ 00:04:05.080 Pranav: Certain, like, meeting notes, yeah, etc, so… Ed.
25 00:04:05.540 ⇒ 00:04:13.449 Pranav: It’s, it’s impressive. What we did, too, is, like, we’re pulling in, basically, these meetings on a daily basis as well. So, like.
26 00:04:13.500 ⇒ 00:04:30.110 Pranav: we have, like, a system in place internally where it’s like, okay, it’s not just, like, ad hoc on request, searching for these notes. We have a system in place where it’s like, every morning, it runs through the previous day’s meetings, it embeds them, and then now when you ask a question, it has all of that context.
27 00:04:30.350 ⇒ 00:04:33.859 Daniel: Sick. And what are you guys piping it into?
28 00:04:34.900 ⇒ 00:04:41.570 Pranav: Yeah, so we have, like, a… we have, like, a data warehouse where we’re just, like, storing the embeddings in GCP.
29 00:04:42.520 ⇒ 00:04:50.130 Daniel: Alright, so just set up a GCP container on this, you’re just dropping all the… but, like, basically transcripts, or is it the summaries?
30 00:04:50.680 ⇒ 00:05:08.029 Pranav: It’s, yeah, the transcripts aren’t on for all meetings. We talked to Adam, and they’re like, we don’t want to necessarily do that. And I don’t think we need to, because those are gonna be a lot of data, right? That’s kind of something we’re balancing. It’s, like, going too far, and then having to manage all that data. We’re just using meeting notes.
31 00:05:08.380 ⇒ 00:05:12.850 Daniel: Sweet. So it’s basically pulling the Gemini meeting notes, dropping them into an MCP,
32 00:05:13.240 ⇒ 00:05:20.469 Daniel: Or, whatever, a warehouse there. And then you guys built a little semantic engine on top of it that just basically tries to say, hey.
33 00:05:21.130 ⇒ 00:05:25.910 Daniel: what are the themes here? I have to say, the email you sent me is pretty impressive, Pranav.
34 00:05:25.910 ⇒ 00:05:30.839 Pranav: Okay, cool, cool. And, formatting’s not fully there yet. Me and, Mustafa…
35 00:05:30.840 ⇒ 00:05:31.450 Daniel: about that.
36 00:05:31.450 ⇒ 00:05:32.969 Pranav: I know, I knew you wouldn’t care about that.
37 00:05:32.970 ⇒ 00:05:39.999 Daniel: I could copy this whole damn thing and put it into Claude and tell it to make me a… like, I don’t care about that. It’s the… it’s the underlying…
38 00:05:40.240 ⇒ 00:05:46.369 Daniel: Like, text reinforcement and, like, that semantic component that just cannot get done.
39 00:05:46.730 ⇒ 00:05:47.270 Pranav: Yeah.
40 00:05:47.270 ⇒ 00:05:49.030 Daniel: Without the full picture.
41 00:05:49.870 ⇒ 00:06:01.859 Pranav: So, okay, I’m glad that we’re starting off there, because I know how you work too, like, you have, like, a bunch of ideas as well. I have a bunch of ideas as well, so I kind of want to talk about, like, okay, this is where we’re at right now.
42 00:06:01.900 ⇒ 00:06:13.009 Pranav: this is kind of what we wanted to do for this project, right? We wanted to detect these themes and then just build an automated, like, process where you get an email sent every Monday, or every Friday, whatever.
43 00:06:13.580 ⇒ 00:06:16.849 Pranav: now I have a few ideas, like, what’s next, you know, it’s…
44 00:06:16.870 ⇒ 00:06:32.719 Pranav: the easy things that come to mind to me, like what we’ve done in the past, is just building actions on top of these items, right? So we showed you, like, kind of the bottlenecks on there. Now there’s probably, like, a certain, like, playbook that you go by whenever you see certain type of issues pop up.
45 00:06:32.720 ⇒ 00:06:45.899 Pranav: Sending Slack messages, coordinating a meeting with, with the leadership to kind of get us back on track, things of that nature. Building agents that, like, run that stuff directly based on this report makes a ton of sense.
46 00:06:46.010 ⇒ 00:06:53.600 Pranav: So that’s one thing. We can kind of go into further depth about that, fully refine that. Another thing is integrating more data sources.
47 00:06:53.740 ⇒ 00:07:01.750 Pranav: I remember in the beginning, way back when we first kind of met, you were talking about how Zendesk and just phone calls, too, have a lot of data.
48 00:07:01.880 ⇒ 00:07:17.360 Pranav: And so, integrating those phone calls, we’ve done that for, like, phone transcripts and, like, for other projects as well. Basically, similar to what we did with, Gemini Meeting Notes, just embedding the transcripts and doing semantic search on top of the transcripts.
49 00:07:17.420 ⇒ 00:07:24.040 Pranav: to just further, kind of, refine that, like, output that I sent to you via email, so…
50 00:07:24.160 ⇒ 00:07:29.900 Pranav: Those are the two, kind of, big, kind of, directions, like, and continuations for this, but I’m curious to see what you think.
51 00:07:31.740 ⇒ 00:07:51.609 Daniel: Two really interesting, really specific use case concepts. The actioning and agent side, I’m not quite ready to touch that yet. Not because that wouldn’t be useful, but because I… I don’t know, I’m just not quite ready to touch that yet. Okay. The phone call and understanding, that’s really like a know-your-customer kind of use case, which is super cool.
52 00:07:53.200 ⇒ 00:08:05.610 Daniel: I might consider what we could do with that, but I actually want to start on a different angle, which is… so as we architect this system, I mean, basically what you guys have done is built a…
53 00:08:05.770 ⇒ 00:08:14.319 Daniel: pipeline, where you recognize a certain context of information in email threads, whatever, you’re just pulling it… I’m assuming you’re pulling it straight out of Google Workspace, right?
54 00:08:14.320 ⇒ 00:08:15.000 Pranav: Yep.
55 00:08:15.000 ⇒ 00:08:32.369 Daniel: So you’re pulling all these meeting notes, you set it up to pull meeting notes, then you build out a meeting notes thing that just drops it in, does a semantic analysis on top of it. The semantic component is where all the prompting and engineering goes on this side, right? How do you spot signals, how do you trend, you know, do this trend and all this kind of stuff?
56 00:08:32.539 ⇒ 00:08:35.260 Daniel: So, two questions in terms of architecture.
57 00:08:35.440 ⇒ 00:08:41.660 Daniel: if… like, like, how do I access this database? So you guys have basically one…
58 00:08:42.110 ⇒ 00:08:45.540 Daniel: tool, one system, what have you put on top of it? What is that?
59 00:08:46.280 ⇒ 00:09:05.619 Pranav: Yeah, so you can kind of think of these as, like, individual, yeah, just like you said, tools. So we have a meeting notes tool. So what that does is that does a semantic search on top of the data warehouse that has all the embeddings of the meeting notes. So yeah, tool that basically can query on top of the meeting notes, semantically.
60 00:09:05.770 ⇒ 00:09:07.579 Pranav: Not just based on keywords.
61 00:09:07.580 ⇒ 00:09:12.609 Daniel: That’s sick. And then… and then this… but… but this tool, is that going to…
62 00:09:14.100 ⇒ 00:09:16.619 Daniel: So we got this data warehouse, we got this tool.
63 00:09:17.740 ⇒ 00:09:29.629 Daniel: Is this gonna be a browser-based tool, right? I go in and can ask it questions, or it sets… you guys have presumably set this up for some type of automated email on a weekly basis, or something like that?
64 00:09:30.390 ⇒ 00:09:47.979 Pranav: Yeah, so it’s set up for automated… so this is one tool of the many tools that we’ve created. So we also check, you know, Google Calendar, we check Slack, we check all these different things. Think of each one of those, like, avenues as individual tools. And then there is an orchestrator that
65 00:09:48.010 ⇒ 00:10:06.049 Pranav: only query, or only, triggers the relevant tool based on the user’s message. So basic… yeah, so that’s… that’s basically how all of these, like, chat interfaces work, like, even, like, if you’re using Cloud Code, too. Like, they have a bunch of tools under the surface that they will only trigger based on
66 00:10:06.270 ⇒ 00:10:09.019 Pranav: What it thinks is relevant based on your request.
67 00:10:09.020 ⇒ 00:10:24.910 Daniel: skill deployment, or whatever else, how they orchestrate all that. Okay, so this orchestration layer, what is that? Like, is that a SaaS platform you guys use? Is that a, like… yeah, I know you sent me the original link, so I’m assuming there’s a new link to access this, right?
68 00:10:24.910 ⇒ 00:10:26.869 Pranav: Yes, yes, I can send that over to you.
69 00:10:26.870 ⇒ 00:10:33.000 Daniel: And then, is that a… some type of open source, or… what is that?
70 00:10:33.410 ⇒ 00:10:41.799 Pranav: Yeah, so there’s actually not a ton of complexity that comes with that. What it really is, is just another…
71 00:10:42.110 ⇒ 00:10:49.170 Pranav: found, like, another just tool, you can think of it, that has the ability to understand each of these other tools. So…
72 00:10:49.270 ⇒ 00:11:00.310 Pranav: you can think of it as it has the metadata for each of these other tools. It knows that for meeting notes, you can… these are the type of conversations that are happening in meeting notes,
73 00:11:00.810 ⇒ 00:11:11.280 Pranav: we also… well, how we built out the system is that we make sure that it goes through most of the different data sources as well. And so, like, one thing, like…
74 00:11:11.410 ⇒ 00:11:26.730 Pranav: think of it, to be put very simply, is that it’s a very, like, intricate, like, system prompt, that has the understanding of each of the tools. It knows, like, what its limitations are,
75 00:11:27.180 ⇒ 00:11:29.150 Pranav: Yeah, does that kind of…
76 00:11:29.150 ⇒ 00:11:30.840 Daniel: In a simple way, what is it?
77 00:11:32.030 ⇒ 00:11:41.549 Pranav: Yeah, so… it… it has all of the… it has all the context of each of the tools, so it’s like a… it’s a prompt, but then it also has…
78 00:11:42.260 ⇒ 00:12:01.210 Pranav: it’s not as simple as a prompt, because a prompt is going to be more undeterministic, right? We also have some determinism in here, where it’s going to specifically call certain tools, on every request, as well. So, like, the meeting notes tool gets called on every single request, because we know it’s extremely information-dense.
79 00:12:02.280 ⇒ 00:12:04.619 Daniel: Yes. Love all of that.
80 00:12:04.850 ⇒ 00:12:09.609 Daniel: I guess what I’m asking is, what platform is the orchestrator? What… what…
81 00:12:09.610 ⇒ 00:12:10.600 Pranav: Oh.
82 00:12:11.440 ⇒ 00:12:17.880 Pranav: Yeah, it’s not, it’s just, a Gemini… Gemini call. It’s an Gemini API call, yeah.
83 00:12:17.880 ⇒ 00:12:18.510 Daniel: Cool.
84 00:12:18.940 ⇒ 00:12:27.090 Daniel: Well, it’s pretty sweet. You know, this is… this is a project I’m probably gonna play around with. This may be parked for a moment here.
85 00:12:27.500 ⇒ 00:12:27.850 Pranav: Okay.
86 00:12:27.850 ⇒ 00:12:31.360 Daniel: you can appreciate. But, in essence, when… when…
87 00:12:31.930 ⇒ 00:12:48.820 Daniel: we have this, there’s two things I want to make sure, which is, number one, like, I have the tool, we can access it, I can start asking it questions, you know, play around with this, and it’s… it’s… the context is going to keep piping in. I guess that’s my first question. So is this set up just to auto-populate?
88 00:12:49.510 ⇒ 00:12:52.350 Pranav: Yeah, so right now, what the…
89 00:12:52.400 ⇒ 00:13:02.569 Pranav: the email that I sent you, and I think the link that you have is for our production interface. Right now, all of the latest changes are on dev, while we just continue doing a little bit more QA.
90 00:13:02.580 ⇒ 00:13:17.729 Pranav: We’re still, like, yeah, making sure that pipeline is pulling in the data on a weekly basis. We still need to do a little bit more testing on that. Or, sorry, on a daily basis. And then the email generation part is, like, something we kind of built a V1 of, like, 2 days ago.
91 00:13:17.730 ⇒ 00:13:24.739 Pranav: And so that part, you know, we still need to do a little bit more QA testing on. Since that’s gonna be running weekly, like.
92 00:13:24.740 ⇒ 00:13:31.450 Pranav: I want to spend, probably, like, next week, we can make sure that that’s… that’s completely good to go.
93 00:13:31.450 ⇒ 00:13:45.740 Pranav: So, but that’s where we’re at. Like, now it’s really, like, we’ve shown you, like, the output, you’re pretty happy with it, now we just need to push things into production, and the link that you have, it should still work. I will send you the dev link for right now, so you can mess around with it a little bit more.
94 00:13:47.000 ⇒ 00:14:02.899 Daniel: Send me the prod link, when you guys get through it. Okay, sure. I’m excited to see how this workflow works. I want to start putting into action after you guys have, you know, had a chance to get some FP. Again, it’s actually… I was impressed by how
95 00:14:03.710 ⇒ 00:14:06.749 Daniel: That what you… whatever you guys did to change it from…
96 00:14:07.080 ⇒ 00:14:10.840 Daniel: to really hit that semantic output makes this…
97 00:14:11.150 ⇒ 00:14:14.079 Daniel: richer than I can grab out of my clot at this point.
98 00:14:15.390 ⇒ 00:14:16.110 Pranav: Yeah, yeah.
99 00:14:16.110 ⇒ 00:14:29.799 Daniel: That’s kind of the bar I was looking for, which is, I can go ask Claude what’s going on in Slack based on the access I have, right? Yeah. But getting that semantic layer across the company with these meetings was really the blind spot, and this hit a lot of those. It was pretty cool.
100 00:14:30.370 ⇒ 00:14:46.279 Pranav: Yeah, and so the reason why Claude can’t do that is because it doesn’t do semantic search in the same way that we did it. It is not having an ETL job run every single day to process every single document within the organization.
101 00:14:46.380 ⇒ 00:15:01.749 Pranav: the best way that I can explain, like, the way it does without getting into super technical is that every file has some metadata, right? It has, like, a, like, every doc, you know, might have, like, a title, it might have, like, who are the collaborators.
102 00:15:01.750 ⇒ 00:15:07.590 Pranav: That’s the only context that Claude will look at to assess whether it’s relevant or irrelevant information.
103 00:15:07.960 ⇒ 00:15:11.400 Pranav: So, it’s like… it’s a level of abstraction that…
104 00:15:11.400 ⇒ 00:15:19.790 Daniel: Yeah, it’s a hierarchical search function, right? Is this title applicable? Yes. Okay, are the collaborators matching? Yes. Okay, then we’ll read the text context.
105 00:15:19.790 ⇒ 00:15:21.089 Pranav: Right. But what we’ve noticed…
106 00:15:21.090 ⇒ 00:15:24.790 Daniel: Set it up to look for a thematic, like, assess theme.
107 00:15:25.010 ⇒ 00:15:29.010 Daniel: dive deep into text, or… or… actually, tell me what to do. I’m guessing.
108 00:15:29.010 ⇒ 00:15:34.270 Pranav: Yeah, yeah. So, how this semantic search works is basically,
109 00:15:34.490 ⇒ 00:15:46.519 Pranav: It’s… it’s like complex linear algebra that basically looks at the prompt and assesses how semantically similar it is to the… each piece of content that’s within every single meeting note.
110 00:15:46.630 ⇒ 00:15:52.680 Pranav: And so, based on, like, it’ll find, with this linear algebra, assess what is the most…
111 00:15:52.820 ⇒ 00:16:05.980 Pranav: after, like, we set a certain bar to say, okay, if it’s under this bar, it’s irrelevant. If it’s over this bar, it is relevant. And, you know, it’s a little bit more complexity there, too. Like, we also prioritize things, we’re pulling in too much information.
112 00:16:05.980 ⇒ 00:16:14.620 Pranav: But basically what that allows us to do, which Claude can’t do, is we’re searching across every single word in the meeting notes.
113 00:16:14.620 ⇒ 00:16:25.699 Pranav: Right? Claude isn’t gonna do that. It’s not even gonna read certain documents, because it’ll just say, okay, based on the meeting titles… I mean, sorry, based on the document titles, we’re only gonna look at these ones.
114 00:16:26.800 ⇒ 00:16:28.080 Daniel: That’s pretty sweet, man.
115 00:16:28.300 ⇒ 00:16:33.009 Daniel: I’m really pleased with that output. That was a great overall email, I love that.
116 00:16:33.120 ⇒ 00:16:33.990 Pranav: Cool.
117 00:16:35.100 ⇒ 00:16:48.420 Daniel: So, sounds like go for… so, appreciate it, that’s awesome. Also, great context. It’s kind of fun to see, like, you know, just overall, why certain things work and why certain things don’t. Yeah.
118 00:16:48.470 ⇒ 00:17:07.500 Daniel: the… the question I’ve got for you, so… so to kind of wrap up that standpoint, basically what you guys are going to do is say, okay, I’m gonna package this, you know, this prompt and everything else in a… in a deliverable. It’s gonna live on some URL where I go access this thing, right? And then it’ll basically access all this, we’ve got the data piping in.
119 00:17:07.609 ⇒ 00:17:21.589 Daniel: expanding beyond that, adding new tools, adding new data layers, all that kind of stuff, we’ll explore how we want to approach that. If we want to do this similar type function with, like, what your customers are saying right now type of analysis, that’d be a really cool project, too.
120 00:17:21.890 ⇒ 00:17:29.130 Daniel: But I want to get this one to the point where I’ve got it, like, in prod and workable, and play around with it, and see what the power of that tool’s gonna be.
121 00:17:29.500 ⇒ 00:17:34.409 Pranav: Cool. Yeah, let’s, we’re very close to that, yeah. Sweet.
122 00:17:34.620 ⇒ 00:17:39.009 Daniel: Well, this is great stuff. I love this. This was awesome. The fact that it’s pulling… and…
123 00:17:39.450 ⇒ 00:17:56.239 Daniel: after going through everything, like, I realized the real distillation of a lot of this is, in fact, meeting notes. Yeah. Like, that’s almost what’s missing… that’s the piece that Claude is not picking up that provides so much context into pushing those themes.
124 00:17:56.820 ⇒ 00:18:01.999 Pranav: Yeah. I mean, that’s what we noticed internally, too, like, they’re…
125 00:18:02.210 ⇒ 00:18:06.460 Pranav: it’s all just kind of a balancing act, right? Claude cannot…
126 00:18:06.470 ⇒ 00:18:15.469 Pranav: build a system that is going to allow for the type of semantic search that we’re building with this. Just because it’s, like, a one-size-fits-all.
127 00:18:15.470 ⇒ 00:18:27.650 Pranav: Type of system, and it’s meant to be more lightweight. Building something more heavyweight like this, like, requires a lot more development, and it requires a little bit more management up until you get to production.
128 00:18:27.680 ⇒ 00:18:34.119 Pranav: So it makes sense to me. There is a use case for Cloud Code.
129 00:18:34.290 ⇒ 00:18:38.550 Pranav: However, yeah, like, for this right here, like.
130 00:18:38.970 ⇒ 00:18:54.050 Pranav: It’s good to hear from you, too, because we… I don’t know if you saw also Sam’s analysis of, like, what your cloud co-work looks like compared to what the command center was doing, and you can see, like, the increase in tool calls, you can see, like, how it’s pulling in even more meeting notes,
131 00:18:54.190 ⇒ 00:18:58.640 Pranav: You’re just getting more, more detailed output.
132 00:18:59.350 ⇒ 00:19:07.740 Daniel: Well, it’s actually, ironically… yes, it’s more detailed, it’s actually more summarized, and it’s more…
133 00:19:07.930 ⇒ 00:19:16.879 Daniel: semantic in nature. Like, this isn’t telling me, hey, this thing happened, it’s telling me I’m seeing this thing happen multiple times. And that’s the context that’s really…
134 00:19:17.500 ⇒ 00:19:18.170 Pranav: Yeah.
135 00:19:18.950 ⇒ 00:19:33.770 Daniel: Sweet, man. Well, I’m stoked about it. Really excited to see the prod version. I don’t know where we go after this, but really, it was about creating the core data layer, having that access point, and just having a tool that produces a semantic output, and it was pretty fun to walk through that exercise.
136 00:19:33.990 ⇒ 00:19:36.169 Pranav: Nice. Cool, yeah, this is fun for us, too.
137 00:19:36.170 ⇒ 00:19:50.220 Daniel: This is a, yeah. Somebody should be doing this. I was curious, I’ve been talking to those Worklytics guys, and they’re like, yeah, we’re looking to see how we can deploy, like, some semantic logic into these types of things. It’s pretty cool.
138 00:19:50.630 ⇒ 00:20:05.860 Pranav: Yeah, no, I mean, this was something that, we’ve talked about a lot internally as well, where it’s like, just having line of sight of, like, what is going on across the company, it’s like, it’s not an Eden problem, it’s a every-organization problem. And so…
139 00:20:05.880 ⇒ 00:20:25.399 Pranav: it requires us to think more critically here, too, is like, I think there’s a lot of themes across all organizations where data is usually going to exist in one certain location. I think we found for y’all, it’s meeting notes. For some people, it might be, you know, Slack DMs. For other people, you know, it could be elsewhere, but
140 00:20:25.570 ⇒ 00:20:42.279 Pranav: Yeah, it’s really about finding that, and then just kind of making sure to focus on that. We showed you that first output, right? Where it was not triggering meeting notes every time, and it was not hitting the bar. It was doing a good job if things were only in Slack, but if things were meeting high.
141 00:20:42.280 ⇒ 00:20:46.190 Pranav: It was… it was not doing a good job of understanding those issues.
142 00:20:46.190 ⇒ 00:21:09.940 Daniel: also building the wrong tool to set the semantic themes, and so the meeting… it’s almost like we’ve re-architected that and say, okay, meeting notes are where we need to find these themes, and then drive them to find other details across the org, which is really cool. Because the interesting piece is, with this output now, I can actually take this, drop into Claude, and say what conversations are happening relevant to Slack, if I want to go to that next step. That’s a feasible possibility.
143 00:21:09.940 ⇒ 00:21:13.310 Daniel: Because what I was really missing was the highlight themes.
144 00:21:14.080 ⇒ 00:21:15.449 Pranav: Yes, yeah.
145 00:21:15.660 ⇒ 00:21:16.049 Daniel: That’s pretty cool.
146 00:21:16.050 ⇒ 00:21:26.629 Pranav: So, one other thing, too, that I wanted to bring up with you is, like, how do you like this as, like, a standalone interface? I know we talked about, you know, this is just your command center,
147 00:21:26.770 ⇒ 00:21:39.830 Pranav: But do you want to integrate it into Slack? Because, like, I’m just wondering, like, that’s where you’re doing all your work. If you’re gonna be tagging other people anyways, like, and, you know, to do certain things, do you want just access within Slack itself? .
148 00:21:39.830 ⇒ 00:21:48.269 Daniel: I don’t know that I care as much, because basically, here’s what’s actually gonna happen in the workflow, right? Yeah. I’m gonna find a semantic theme on this.
149 00:21:49.570 ⇒ 00:21:50.250 Daniel: like…
150 00:21:50.630 ⇒ 00:21:58.030 Daniel: none of these things, like, I can’t solve, you know, interconnected challenges with legacy systems with a Slack note.
151 00:21:58.350 ⇒ 00:22:02.810 Daniel: Like, that’s, like, something I need to… restructure…
152 00:22:03.280 ⇒ 00:22:12.559 Daniel: my team’s cadence around. Like, it’s not… none of these are like, hey, I can go tag, you know, Adam Poma and go fix AI integrations in the company.
153 00:22:12.840 ⇒ 00:22:14.160 Pranav: Okay, yeah, yeah.
154 00:22:14.160 ⇒ 00:22:16.629 Daniel: I can’t, like…
155 00:22:17.210 ⇒ 00:22:25.490 Daniel: I can do that with what Claude was producing out, right? Hey, you know, there’s a hiring gap in Brad’s team. Okay, Brad, where are we at with this role?
156 00:22:25.870 ⇒ 00:22:26.440 Pranav: Yeah.
157 00:22:26.440 ⇒ 00:22:37.090 Daniel: Like, that’s super easy to do, but what I really wanted to get was what are the big talk-to-task things, like unresolved junction integrations, you know, into Health OS work?
158 00:22:37.110 ⇒ 00:22:47.520 Daniel: Yeah, that’s, like, probably a 60-90 day project, but I’m really glad that this is picking up, that we have a lot of meetings about this, because that’s exactly what should be happening at this moment.
159 00:22:47.610 ⇒ 00:23:06.770 Daniel: I’d be worried if there’s, you know, one of the things I’m looking at here is we’re, you know, focused really heavily on this affiliate, you know, Martech project we’re working on. It’s not picking up that theme, right? Which means to me that a lot of this is happening in blind corners with teams that aren’t actually integrating themselves back into the team.
160 00:23:06.770 ⇒ 00:23:15.930 Daniel: So, I can be… like, my brain can do a lot, too. Like, I can look at this and see, hey, it was really weird I didn’t see that priority on here, we probably need to get a company meeting to figure out.
161 00:23:15.930 ⇒ 00:23:16.340 Pranav: Yep.
162 00:23:16.340 ⇒ 00:23:19.239 Daniel: want to move that forward. So, this is excellent deliverable.
163 00:23:19.690 ⇒ 00:23:28.339 Pranav: Cool. Yeah, I mean, that makes a lot of sense to me. I think this is something you need to spend, like, probably minutes, if not, you know, a half hour, hour, kind of digesting.
164 00:23:28.340 ⇒ 00:23:36.339 Daniel: weekend thinking about, ugh, that really is a big blocker. I, like, full… mandatory full-body photo requirement for patients. Okay, like…
165 00:23:36.350 ⇒ 00:23:49.140 Daniel: our doctor’s network is legally requiring… saying they legally require it. We’re hearing elsewhere that other doctors’ networks aren’t requiring it. Like, that’s a big project. Like, it should be on this list, you know what I mean?
166 00:23:49.140 ⇒ 00:23:58.659 Pranav: Right, right. Yeah, I see what you’re saying, yeah. That Slack integration is, like, you would want a Slack integration if you’re just like, okay, delegate immediately, delegate immediately.
167 00:23:58.660 ⇒ 00:24:17.630 Daniel: Like, you know, like, I asked Claude to pull my to-dos, and it’s like, okay, I have a to-do to talk to Caitlin about getting the, you know, whatever, the legit script approval for EHC done. Okay, well, that’s an easy tag. I don’t have to… but that’s not a big theme of the company, that’s a two-minute exercise.
168 00:24:17.630 ⇒ 00:24:18.330 Pranav: Right.
169 00:24:18.620 ⇒ 00:24:22.010 Daniel: These are big company themes, which is exactly what I wanted to see.
170 00:24:22.470 ⇒ 00:24:23.040 Pranav: Yeah.
171 00:24:23.610 ⇒ 00:24:27.190 Daniel: Cool. This is great for now. You guys have some cool shit you can do over there.
172 00:24:27.370 ⇒ 00:24:32.950 Pranav: Yeah, yeah. Well, no, this is great. I’m glad we got to sync up, too.
173 00:24:33.080 ⇒ 00:24:39.760 Pranav: Yeah, so next week, I think we’ll have same meeting, same time. We’ll, keep giving you updates on just, like.
174 00:24:40.030 ⇒ 00:24:41.620 Pranav: Just pushing into production.
175 00:24:41.620 ⇒ 00:24:47.029 Daniel: It’s gonna be tough for me. I’m… I’m actually moving Wednesday through Friday.
176 00:24:47.030 ⇒ 00:24:47.640 Pranav: Okay.
177 00:24:47.640 ⇒ 00:24:59.439 Daniel: So, if you want, we can do two things. We can move the date forward to meet, like, Monday or Tuesday. My schedule’s kind of backdating that. Or, we can meet…
178 00:24:59.630 ⇒ 00:25:03.349 Daniel: prob… maybe that following week as I get settled in San Francisco.
179 00:25:03.750 ⇒ 00:25:06.929 Pranav: Yeah, that probably works better. I’ll, I’ll move that meeting to Monday.
180 00:25:07.180 ⇒ 00:25:08.900 Daniel: Okay, cool.
181 00:25:09.540 ⇒ 00:25:11.670 Daniel: Cool. Alright, man. Appreciate it, Pranav.
182 00:25:11.670 ⇒ 00:25:14.639 Pranav: Yeah, appreciate it. Thanks, thanks. Have a… have a good move.
183 00:25:14.950 ⇒ 00:25:17.669 Daniel: Me too. You have a good weekend.
184 00:25:17.670 ⇒ 00:25:19.559 Pranav: Appreciate it. See ya.