May 1, 2026

# Transcript: 2026-05-01

> 2 time blocks from 8:01 AM to 8:48 AM

---

### Weather and casual life complaints
**8:01 AM - 8:15 AM PDT** | *casual*

**Microphone:**
Hey, come on now. It's just dealing with the weather changes. It's hot, it's cold, it's humid, it's freezing. There's a million mosquitoes in your front hallway. There's a bird that just shit on your deck. It's just a mix.

It felt like a very volatile couple of months just generally everywhere I've gone.

Yeah, so down here, I think this is everywhere across America. El Niño I think is over and now La Niña is coming back, or the opposite—like La Niña died and El Niño is coming. So that's why we're getting atmospheric rivers, or whatever they call it.

It's nice because we don't get enough rain down here. I think I'm bred to live in Seattle, much to my children's chagrin. It's a bummer when we don't have the rain.

Hey, Todd, you made it back.

Yeah, I did. Sorry, the Wi-Fi's pretty rough here. It wasn't too bad overall—only delays lately.

That's good. Nice. Cool. Okay, Kim, why don't you kick this off?

Sure thing. Yesterday I started with a quick PR review with Michael going through open items. Then I wrapped up some minor fixes on the PubSub PR and retested that after confirming the changes we implemented. Similarly, I tested the analytics refresh again, made sure specs were passing, and wrote up a ticket for the issue with the relationship leads export. I got that merged as well, then spent the rest of the day working through the final touches on the ReliAi search view filters work. This has gone pretty well—it's done. I'll open a PR this morning to get that in. I just saw the approval on the RallyEye filters PR, so I'll get that in as well. Once all that's merged and deployed this morning, things should be good. That's me.

How about you, Todd?

Yeah, I did some final testing and cleanup of that cleanup PR in 2.8. I'll probably put that into the dev branch and maybe start a new branch as my new base. That way, any fixes, updates, or changes we need to make will be in that 2.8 dev branch as we go through it.

So I'll deal with anything we find and see how things play out with all the different configurations, accounting for both the user and system impacts. If we can figure out a way to do it properly with our own organization, great—if not, we'll just roll with the punches. Because as we know, things are not always smooth sailing with Salesforce.

So basically we're reacting to that. Then I was going through Linear looking at some of the old tickets that initially started all this work and planning the next steps in Salesforce. Basically, I need to start bringing some stuff over to Matt to get closer to parity. I know there's a few cleanups in there. What I might do is leverage that named param we created for which server instance you're running.

The way I've got the dev set up, I think I could pull that off easily now. Instead of having three configs, I'd have just one that you edit in place. That would probably be my first step—a nice way to make progress while we wait to see how the testing shakes out, without risking any major issues with Salesforce. That's me.

Cool. Yesterday I have to look at the calendar. I barely remember Tuesday. So I met with David and some others, specifically for the Swarm call.

We met with Swarm's CEO and CTO. It was pretty standard for a first meeting, I'd say. We haven't all discussed it together yet. Now that David's not on the call, I'll speak frankly—I don't think anyone wants to go work for Swarm. I know I don't, and I don't think David does or any of you do either.

I think we're very, very close to actually getting this thing off the ground. What I took away from it is that maybe there's a way where they could give us money and we give them some of our revenue by using their data—kind of a partnership where we don't have to pass through data for everything, and instead we share it with them. There are some potentially interesting avenues there.

There was a fair amount of focus on the agent and Salesforce. When the CTO asked how I'd sum up Salesforce package development, I said my engineers would describe it as "let me scream into the microphone for 30 seconds and then I'll be able to talk about it." That pretty much captures it.

Interesting fact—they have a 1.5 terabyte in-memory graph database running in Neo4j, which is how they operate so fast. They're paying five thousand dollars per hour for server infrastructure. I looked it up—that's more than we pay for all our servers combined in a month. So that explains a lot. And they're probably burning through a ton of VC cash.

That's definitely a crazy strategy, but it's such a cool idea, right? Like, put a whole graph database in memory—it's going to be fucking fast. And if you get a lot of people using it, it can start to scale.

But this is also the problem with startups that have a lot of people from Uber or DoorDash or whatnot, where they just build the most complex architectures. When CK was here, for instance, we built an architecture. Nobody said it wasn't a good one, but it was such a big architecture for such a small company. So it makes sense if you can get a big raise and a lot of money.

Then I was doing some PR stuff, some cleanup stuff, and then I was working on the next version of the thing we talked about yesterday. Actually, it's a showroom now. You guys had some good feedback at the time, so let's see it. I shared it last night with David. I actually also showed him an example of this I want to do on the connector.

But you know, I think it would be beneficial for us for rolling out some tours like this. We don't have to pay for Pendo to do it. But the real thing is that we're going to have to move to this again, similar to what I showed yesterday. Really, like the whole point is this thing is always running. You don't have to do stuff, but here are the things you do need to do something on. So frankly, limiting the amount of buttons and stuff like that. That's where I was working with David on some of that.

And today I'm trying to take all of these things—like what does the concierge do? Some of this we've already built, like sending sync requests or mapping networks and stuff like that. But I'm trying to put together a list of all tools we want to build into the agent side of the concierge. I'm trying to differentiate the two. The agent is the thing that works internally to fill out data, do research, and so forth. I want us—I thought about this, I think he's on board—I want us when we talk about the thing that interacts with connectors, because that's the one that should be like a concierge. Kind of groveling, very supportive, very helpful, not getting in your way, giving you what you need before you realize you need it.

So I'm trying to figure out what the things the agent should be doing in the background to do research and stuff like that. I'm going to put together a list for us to share later today. Then on Monday at the team meeting, I think we'll talk about that and start breaking it up, because there's going to be somebody who wants to build it.

That's what we were working on today. And then target companies—I still have some stuff to go back to there and settings. I'm kind of feeling like maybe I just wrap settings into target companies at this point, because there's some stuff that's integrated there, and it might just make it easier and faster to roll it all out. So I'm thinking about it. I think that makes sense.

Then I got in a little argument with the SOC 2 auditor. I think I mentioned she was asking for things from 2023, and I was like, "How is that in 2025? You crazy person?" They came back yesterday and signed off, and then her boss was like, "Yeah, my God, Jack, that's—we've already passed off to the auditor. Like, that's crazy. We're gonna get you a review next week." So we could be in a good place there.

And we killed the Demo Environment yesterday. Then I will find some time. The thing we did to kill the target company's aggressive fan-out with like one word, and now we want to do that to people because it seems to be working for target companies, and the people data indexing is eating up a lot of energy and a lot of jobs.

There's that, and then Cam, I'll just say I did push up the stupidest, dumbest PR in the world. It's like a README-type change. I want to clean up how the prototypes are done, and I pushed that up yesterday into SmallWorld Prototypes. So that means as I'm working, if you go to prototypes.smallworld and you log in through Google, they're just trying to be super transparent. Like, anything I'm working on, you can see it there. So if you have thoughts, or you just want to see what I've been working on, in that regard, you can always kind of see it there. Your prototypes look awesome.

Yeah, we're—I mean, to be honest, I think David likes having these because, you know, selling to clients is like, "Hey, we can do this tomorrow." Selling to investors is like selling the vision. And so being on the show where we're going is really, really helpful for him.

So yeah, it's interesting if you look at the relationship map V2. Like, for instance, it's funny how it kind of thinks through conversations and feedback. But it's interesting to kind of work with it and be like, "Yeah, that's not what I said, man."

Or he'll over-index for David and he'll be like, "David specifically said we should call this." While you were sleeping, I was like, "David didn't say that. Yeah, David referred to it as that, but he didn't specifically say it needed to be called that." So it's funny how it indexes on people.

Okay, so with that, anybody else have anything they need right now? Cool. Let me know how you guys get into testing today and how that goes. If there's anything we need to do on my side or so forth and so on as it comes to Salesforce, let me know, okay? Copy that. Cool. Okay, guys. Thank you very much, and I'll talk to you throughout the day. All right.

**System Audio:**
Hey, Cam, I know. Okay, we're getting a lot of rain, which we need, but it's just kicking my ass with the weather changes. It's hot, it's cold, it's humid, it's freezing. There's a million mosquitoes in your front hallway. There's a bird that just throws debris on your deck. Like, it's just a mix.

Yeah, so down here, and I think this is everywhere across America. El Niño, I think, is over, and now La Niña's coming back. Or the opposite - La Niña died, and now El Niño's coming. So that's why we're getting atmospheric rivers or whatever they call it.

So I mean, it's nice because we don't get enough rain. I think I'm bred to live in Seattle, much to my children's chagrin. But it's a bummer when we don't have the rain, but it's nice.

Hey, Todd, you made it back.

Todd Leffmanislein.

Oh, I did. Yeah. It wasn't too bad. We had slight delays.

That's good. Nice. Okay. Cool. Okay, Kim, why don't you kick us off?

Final testing, final cleanup. So that cleanup PR that is in 2.8 - I'll probably just put that into the dev branch and maybe start a new branch. That would be like my new base as we go forward, so that can be, you know, any updates or changes that need to be made for it as we go through it will be in that 2.8 dev branch.

And then I'll just deal with anything that we find and see how it works. How things play out with all the different configurations and the user - like if we can figure a way to do it "in quotes" with our own organization and kind of just roll with the punches.

Because as we know, things are not always smooth sailing with Salesforce. So basically we're reacting to that.

And then I was going through Linear and looking at some of the old tickets. I initially started all this work, looking at the next steps in Salesforce, which is basically kind of starting to bring some stuff over to get closer to parity.

I know there's a few cleanups in there. And that - like, especially what I might do is - I know there was a named param that we have with which server instance you're running. And the way I've got the dev set up, I think I can pull that off pretty easily now. Instead of having three configs, have just one that you edit in place.

So that would be my first thing - try to get that working. That would become a nice placeholder to still make progress but kind of wait on any major blowups in testing.

So Salesforce again, that's me.

Cool. Yesterday - I have to look at my calendar. I barely remember Tuesday.

So I know I met with David a bunch of times specifically for the Swarm call. So we did meet with the Swarm CEO and CTO. It was a pretty standard first date, I would say. In terms of what we discussed, we didn't all talk about everything, but I'll try to speak frankly: I don't think anybody wants to go work for Swarm, and I don't think he wants to either, nor do I. I don't think any of you guys do either.

I think we're very, very close to actually getting this thing off the ground. And so, you know, we're just looking at options. I kind of took away from the meeting - it's like maybe there's a way where they can give us some money and we can give them some of our revenue by using their data and like do it like a partnership like that where we don't go through LIMA data. They didn't say for some things we give to them instead, blah blah blah. So that was, you know, that was my two ways - there's some avenues there I think that were potentially interesting.

I think my team is going to have a fair amount of focus on the agent and Salesforce. The CTO - I've heard a lot of interesting things about developing Salesforce packages. The engineers would sum it up as, "Let me scream into the microphone for 30 seconds and then I'll be able to talk about it."

Interesting fact: they have a 1.5 terabyte in-memory graph database running in Neo4j, which is how they're able to operate so quickly. So it's amazing how fast they can do something like that.

I was like, they're paying $5,000 per hour for that server. I looked it up. That's more than we pay for all of our servers across a month, Dave. I should chill out.

So that explains a lot. Also, I have to imagine they're burning through a lot of VC cash. Interesting conversation there.

It's such a cool idea though, right? Like, you know, put a whole graph database in memory. It's going to be fucking fast, and if you can get a lot of people using it then it can maybe start to scale. But this is also the problem with startups that have people from Uber or DoorDash or whatnot, where they just build the most complex architectures. It's not unlike what happened when CK was here - which is like, you know, building an architecture. Nobody would suggest the architecture wasn't a good one, but it was such a big architecture for such a small company.

So it makes sense if you can hit it big and raise a lot of money. But anyhow, that was pretty cool.

Then I was doing PR stuff and some cleanup stuff, and then I was working on the next version of the thing we talked about yesterday and I can actually show right now because you guys had some good feedback at the time. Maybe you'd like to see it. I shared it last night with David.

I actually also showed him an example of this approach. I want to do this on the connector dashboard, but I think it would be beneficial for us for rolling out some tours like this. We don't have to pay for Pendo to do it.

But the real thing is moving to this approach - similar to what I showed yesterday, but really, the whole point is this thing is always running. You don't have to do anything, but here's what you do need to do something about.

So really, frankly, limiting the amount of buttons and stuff like that is where I was working with David on some of that. And today I'm now trying to take all of these things here—like what does the concierge do? Some of this we've already built, like sending strength requests or mapping networks and stuff like that. But I'm trying to put together a list of all of the things we want to build into the agent side of the concierge. It says concierge here, but like the agent side of the concierge—I'm trying to differentiate the two. The agent being the thing that works internally to fill out data, do research, and so forth.

I want us—David, I've talked about this, I think he's on board. I want us, when we talk about the concierge, to talk about the things that we're doing that interact with connectors because that's the one that should be like a concierge—kind of groveling, like very supportive, very helpful, not getting in your way, getting you what you need before you realize you need it, et cetera. And so trying to figure out what streams the agent should be doing in the background to do research and stuff like that. So I'm going to put together a list for us to share later today. And then on Monday at the team meeting, I think we'll talk about that and start breaking things up because there's going to be some new ones to build. But, you know, that will come next. So just a heads up on that. That's what I'll be working on today.

And then target companies—I still have some stuff to go back to there and settings. And I'm kind of feeling like maybe I just wrap settings into target companies at this point just because there's some stuff that's integrated there and it might just make it easier and faster to roll it all out. So I'm thinking about that.

And then I got in a little argument with the SOC 2 auditor. I think I mentioned she was asking for things from 2023. And I was like, how is that in 2025, you crazy person? She came back yesterday and signed off, and then her boss was like, yeah Michael, I checked—that's true, we've already passed it off to the auditor. Like, that's crazy. We're gonna get you a review next week. So I think it would be a good place there.

And we killed demo environment yesterday. Oh, and then I will find some time. The thing we did came through—the target companies aggressive fan out with one word. And now I want to do that to people because it seems to be working for target companies, and the people re-index is eating up a lot of energy, so a lot of jobs. Okay, so there's that.

And then Cam, I'll just say, I did push up the stupidest PR in the world. It's like a README type change. I want to clean up how the prototypes are done. And I pushed that up yesterday into small world prototypes. So that means as I'm working, if you go to prototypes.smallworld and you log in through Google, I'm just trying to be super transparent. Like anything I'm working on, you can see it there. So if you have thoughts or just want to see what I've been working on in that regard, you can always kind of see it there.

Yeah, we're—I mean, to be honest, I think David likes having these because he's, you know, selling to clients. It's like, hey, we can do this tomorrow. Selling to investors is like selling the vision. And so being able to show where we're going is really, really helpful for him. So, anyhow, it's interesting.

If you look at the relationship map, V2, like, for instance, it just—some of the stuff is funny. It's funny how Claude kind of thinks through conversations and feedback, but it's, you know, it's interesting to kind of work with it and be like, yeah, that's not what I said, man. Or he'll over-index for David and he'll be like, David specifically said we should call this while you were sleeping. I was like, David didn't say that. David does refer to it as that, but he didn't specifically say it needed to be called that. So it's funny how it indexes on people.

So, does anybody else have anything they need right now? Cool. Let me know how the testing goes today, if there's anything we need to do on my side or, you know, so forth and so on as far as Salesforce goes. Let me know. Okay, cool. Okay, guys. Thank you very much. And I'll talk to you throughout the day. Bye.

### Casual chat about QA work
**8:19 AM - 8:48 AM PDT** | *casual*

**Microphone:**
Here we go. I'm about as good as I was 20 minutes ago. Unfortunately, I had a break in the past 20 minutes—killed my dog, broke my TV, punched myself in the face, stole money. TGIF.

How's it going? I know you've had a lot of QA. Is it frustrating to do QA? Is it annoying? Is it depressing? Is it exhilarating?

I don't think I would call it depressing. After prolonged periods of problem solving, I actually sort of like having focused, closed problem sets. You know, like, I just need to make sure that things are working as expected, as opposed to let me design this, then design the validation criteria, then do everything you need to do to build something. So I don't despise QA or anything, and I sometimes enjoy the variety in my job.

I've been working on the search view filters, obviously, and I like how it works. I can show you right now. It feels good, generally speaking. I'll show some examples here.

Can I ask—could you just show me? You're very strong. Why did it only say "strong" if you have "average" to "very strong" selected? Just say "average or above."

Ah, okay, I didn't understand that.

There's some stuff I could make the argument about, and there's basic if-then logic involved. David's probably gonna suggest something about this. I wish it had like a ticket to do the front-end search view filter stuff.

Yeah, that'd be good. Yeah, but anyways, I think that looks great.

Yeah, and then I think it's fine the way that it is. I might want to make a minor change to the SmallWorld LLM worker because it's saying things like "show me in the SmallWorld network at my target companies." I just want to make sure that doesn't result in it searching for a company named "target company" or "SmallWorld Network" or something like that. But I haven't seen that, so I'm just trying to be defensive.

No, that's good. That's great. Yeah, I mean, that would be a win if we can get that out next week. Then that's another win. We can have the filters, and then we can start deprecating the old requester dashboards. That'll be good.

I was a little sad to see David so caught up on the new Real AI and the filters. He seemed to be expecting more?

No, I thought we were pretty clear.

David, I hear you, and I ran into that similarly, funny enough, this morning. I sent David the link to the thing I showed you, and he said, "This is really cool, you know? Can we run this on four or five accounts and see what a simplistic version could return?" I didn't respond at that time—it was like five in the morning. But my response was like, "David, I did a design. I painted a picture. Like, if you painted a picture of her car, it can't drive, right? You know that. And, like, that's the point."

It's funny, because I literally just said this one. But just because David says something doesn't mean we do it. Just because David says that something should be done doesn't mean that it is done. Likewise, just because David is expecting something doesn't mean he always gets it, for various reasons. We try sometimes, but the challenge is he doesn't understand the custom building stuff and how long it takes. He doesn't both understand nor delve into the weeds enough to understand that.

So you're working on these filters. You're not also working on the agent?

No. Well, because the filters, once you search, you'll do other stuff with the agent.

Yeah, like just because I bought a house doesn't mean I have kids now. There's a process, right? So just keep that in mind. It's not a—with him especially, it's not like he just gets ahead of himself and ends up communicating in a little bit of a clunky way. That's all I'm saying. I appreciate his forward-looking nature sometimes, certainly, but it can be disheartening occasionally. But yeah, I'll take it to heart.

Yeah, I hear you. That is hard.

So that's—I'm glad to see that coming on now.

Yeah, I think it'll obviously go open for review this morning. I'm about to merge the Rallyi filters chain. There's just a Greptile comment that I thought would be good to address. And then, yeah, I'm assuming a nice smooth transition into Salesforce while that's open for review. I think that should pack out the rest of the day. I guess probably including some additional testing on the other PRs that you've got open.

Do me a favor. Let me know at the end of the day your honest take on the Salesforce stuff because I'm still a little confused. Or not confused—I'm a little like, some of these problems seem like we're just doing the same problems over and over again, and I'm trying to understand that. Your thoughts there would be helpful.

Yeah, for sure. So there's that. What else? Is there any other much of work I should be mentally preparing for?

Kind of. I'll tell you about it. You want to hear about it?

Oh, yeah, I would love to.

Okay, give me a second. 

Is there a ticket for that? Do you want me to write that?

You can write it. I don't think there's a ticket for it.

Okay, so I'm working on something and I'll show you what your next piece of work is. Hey, it's this! Not the front-end—it's related to this. Actually, I'm gonna switch screens now. I was just trying to be cute. It's this. This is what I'm working on right now, which is I literally took the designs and I listed all the processes that we're kind of showing. And then I was like, "What else could we add?" I even asked, "Claude, do you have any ideas?"

And Claude came up with some interesting ones, like looking at the PR and LinkedIn posts of people to see if they're champions or skeptics of our company. And I was like, that's crazy, which is funny.

Oh, by the way, yesterday in the swarm call, they told us they had a PE fund when they were trying to sell the PE fund as a customer. And the PE fund was like, "Can you, if we have you follow all of our employees or something like that on Instagram, can you look in their phone and identify potential people they know?" And David literally was staring at him and goes, "Well, that's creepy." Yeah. I was like, yeah, PE companies never proving to anybody that they're not the worst people in the world.

Yeah, I'm sure Edward Snowden's having shivers sent down his spine.

So anyhow, I think this is a duo project where I'll work on the front end and you work on the back end. This goes back to what we were talking about a couple weeks ago—about, hey, what do you want to work on? You want to work on data-related stuff, right? Well, most of this is data, but it's also orchestration. And so what I want to do—I just pulled this list here. I haven't gone through it with David yet. I'm gonna run through it with him, and then on Monday I want to sit with you and say, okay, let's put an architecture, which you've already done in some ways. This is the atomic unit of a process that the agent can run, okay? Or a stream or whatever you call it. And I want to sit down and be like, the inputs and outputs are always structured in this way. It's always going to be a worker in Cloudflare or a service or something like that. It's going to have these types of tests. I want you to be a brain-dead cookie-cutter programmer by the time you're done with this list, so to speak.

Not because I want you to be brain-dead or whatever, but because we want to make it like a machine where we can just keep adding processes as we add more data and partners and stuff like that. Also because eventually we'll hopefully be able to hire somebody else, and they would be a junior for you that could literally be someone who just manages things. I know that's what you've always wanted in life. No, I'm kidding.

But the idea is—so David was talking about the swarm yesterday, and he asked how they have so many integrations. I'm like, this is how you do that, David. Because this is what we do at Webby. They have a Polish CTO who hired a bunch of Polish engineers in Poland who are cheap as shit, though not as cheap as they once were. And you literally go, "Hey, Michal, you're the Clay guy" and "Hey, Brennan, you're the Gong guy," and that's all they do. Or maybe they do two or three, but it's kind of how we have someone working on the Chrome extension or the Salesforce thing. That's how you would do it. And it's easier—you have a little bit of money because you just hire the cheap people to do it.

So in theory, eventually we'll build up the team like that. But the goal is, if I start building the interface over here, we need to fill it up with stuff. And most of this stuff—sorry, I'm just gonna share my whole story for a few minutes because I keep switching back and forth with my notes. I'll just hide that.

So most of this stuff is what should show up in theory here. And it's not like a list of all 20 of them, but it's like when certain things are running, it would show up in the UI. For instance, like being able to allow somebody to reply to an intro right from here—I can build that. I mean, you could build it too, but I can build it, design it, do all that stuff. I could do this too, but you've been doing this and you're better at it. You think like an agent does. So this is better suited for you, I think.

I can do the interface, I can do all this stuff, but if we can collaborate on this list and really frame it out—what's the input, what's the output, where's the data stored, how's it handled, et cetera, et cetera—then I think we add this to a page and it's one item running for a week, then the next item next week it's three items, the next week it's ten items. You know, we just add them as we go. I think that'll be sweet.

Yeah. Is that work you're interested in doing?

Oh, definitely. Yeah. I mean, the one thing that comes to my mind is, do we want to start doing scheduled tasks basically?

Yeah, for concierge, yeah.

Yeah, so the whole point—and this is a little different than we originally talked about—and so to be frank, I don't know where the Relay interface is going. I think the Relay interface eventually just moves over to Slack completely. It's still on the website, but it's really Slack. These teams are where the orchestrator just is always running, and so we'd still show the orchestrations over there too. But it's like different use cases, I think. In any case, we'd have the orchestrator running and they turn it on for a deal team.

So maybe—and this goes back to the use case and the pricing David has been working on—maybe you could just run it on a connector and that's free. But if you're doing it for a deal team, it's all of these 20 things, or 10 or 15 or whatever it is, and they're running constantly. But the point is, it's not going one, two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirty-four, fifteen, one, two, three, four. It's these are the 15 things that the agent can do, and based on the information the agent has, it enqueues those processes as it feels fit.

Understood, yeah, that makes sense. Yeah, it's definitely very orchestrator-heavy, but it's also not like, oh, just cron job everything. It's how do we build the supervisory agent that makes reasonable judgments on—hey, we haven't run the map key personnel job in a while, we should run that. Or we just ran the map key personnel judgment, or key personnel process, and that takes three bucks per company to run. So therefore don't run it anytime soon, you know, or something like that.

Yeah, I bet we can do a lot of that with just some simple heuristics and a view basically. Yeah, or materialized view. I'm not exactly sure what makes the most sense for that for our use case, but I'm sure there's just a couple of fields and data points that we can watch, and if they change, it breaks the cache and the orchestration needs to kick off that task and run that tool and refresh the data.

Yeah, I think so. And you know, we're still bickering on words like David's front deal team, and then he was like in a conversation we had, he was like, yeah, so you have your internal deal team and then you have our team, and I was like, we can't call it what they already call it, like we can't do that. He's like, well no, they'll know it's different. Like they won't know it's different. And we've worked with clients for years now who get confused by the fact that we try to use like what you call a lead is not what anybody else in sales calls a lead. So like, let's stop using the same words. He's like, well, it's like a relationship deal to him. Like that's better. And now he's saying relationship map, which is probably the better answer because that's not an overloaded CRM term.

So, you know, I don't know that that's the right name for it, but it's part of it. It maybe should just be target company. Yeah, I mean, even target companies nomenclature, I find to be potentially problematic just because people don't really understand exactly what it means. It's what we call accounts target companies, and there's like the target companies and buying centers are the two SmallWorld vocabulary terms that I think have done us the most damage. But anyways, that's my point.

I agree. It's weird and I'll be honest—and I don't mean this as an insult against David, but I don't mean it in an insulting way—that David isn't a CRM-first kind of salesperson who builds software who runs a company that builds software for people who are CRM-first, is a hurdle.

Yeah, you really have to know the domain you work in. And he knows sales, but he doesn't totally know how CRMs work, is what I've learned.

Yeah, that's—you know, over the last five years, that is shown to be true. So yeah, don't know how you fix that.

I don't know either. It's a little too late to change his ways in the sales process. Anyways. Yeah, I don't know. As we get bigger, we hire a product person too that will help, and I think we'll see.

Cool. So that's kind of what's going on next, right? And hopefully that will be both, you know, interesting and exciting. And it's also something that we can together collaborate on so that we can support getting it done. And not only getting it done, but like making it iterative, which is something we struggle with, have struggled with at times, right? It's like really making sure that we're working this way. It'll be a nice challenge for us.

Yeah, we'll be fine.

Cool, okay. Um, and I actually have a lot to chat about today to be completely honest. Um, anything on your mind you want to chat about?

All right, well, I mean, that was just about it. I got through most of my plate and what's next, so.

Okay, that's good. I've got to look into the security stuff, and I appreciate you pushing it yesterday. I ran it and oddly enough I don't know if it worked.

Oh, it's sent to me. You're talking about the security review? Slack message? It sent it to you?

Yeah, like from my PR review bot. I'll send you a screenshot right now. Here we go. I was gonna mention it but I assumed you knew that it went directly to me.

Oh. Because you used the one in the—you used like the GitHub environment variable. Yeah. Oh, which is set up for your peer review bot. Gotcha. So we could change obviously, but yeah, I'll do that. That's fine.

Um, what I need to figure out and I honestly just don't know much about it is what's going on in this tab here. Like, obviously we have the dependent pod alerts and that's vulnerabilities, but like, I'm pretty sure we want the Brakeman stuff coming in here too, because this is not a code quality thing, but it's like, um, yeah, I don't know. I don't know if it's supposed to show up here too or what, but it is what it is. But I'm glad that that's running. And if it's not messaging you, we've got to get it to message the DevCon channel or something, but that's good.

Cool, okay. And then what else? PR-wise, oh, so yes, I'll push up that other one for the fan-out for people re-indexing. And then the other thing is, I'm gonna try this weekend to find some time to look at the analytics categories because it's so slow. And it's fine that we have to do it for every person—that's thousands of people—but the fact that they each take a few minutes is a problem.

I mean, I think you'll have really big gains just doing the accounts and even just the connectors index and target companies-related stuff.

Yeah, the connector index is a problem because it's doing it for everybody, but if you just do it on the account level, that is same data for everybody, so that's an easy fix. But the reason I want to look at it is it suggests to me that there might be improvements just in the database thing that need to happen, like indexing or something. So I'm going to look at that.

Cool. Yeah, that's really it on my head.

Awesome.

Just keep me posted as those open up and I'll get that cert view PR open, ready to go. I'll pull up that ticket real quick, and then I'll keep you posted on Salesforce so you know what the word is.

The other thing is they didn't respond. I'm going to check. I'm pretty pissed about this Reptile thing where they bill me. Did I tell you? I mentioned it.

But yeah, you said that you had canceled, but—

No, I had it canceled, but I'm going to cancel if they're going to bill us extra.

Extra?

Extra. Yeah, so they want to bill us. And it's me. It's 29 extra dollars.

Because of your reviews?

Because I have 28 flex reviews. I would have assumed I hit that limit way way way before you.

You are hitting that limit. It's your fault.

Because you're like—it's not your fault, but you are the one doing stuff, but it's on my PRs.

I see. I'm the one being slightly more PR prolific, but you're the one who's acting on the PRs more prolifically.

Yeah, I am.

Who does? Thank you.

I mean, I assume they would probably want to make it right.

Uh-huh.

I will talk to you later.

Sounds good. Will do. Talk soon.

Thanks. Bye.

**System Audio:**
All right. Unfortunately, I had a break-in in the past 20 minutes—they killed my dog, broke my TV, punched me in the face, stole money. Yeah.

How's it going? I know you've had a lot of QA. Which is—I don't know. What is it for you? Is it frustrating to do QAs? Annoying? Is it depressing? Is it exhilarating? Real quick, I'll ask, right?

Ah, okay, I didn't miss that. That's fair. Okay. Yeah. That makes it fast.

Yeah. Like, I could make the argument—David's probably going to suggest me at all, but I would make the argument that connector roles is a very inside baseball kind of thing. Yeah. And you're just doing this through basic if-then statements. Yeah, that'd be good. Yeah. I think that looks great.

No, that's good. That's great. Yeah, I mean, that would be a win if we can get that out in the next week, then that's another win. Have the filters, then we can start on the old requester dashboards and such.

So, David, I hear you. Funny enough this morning—I sent David the link to the thing I just showed you and he said, "This is really cool. You know, can we let's run this on four or five accounts and see what a simplistic version could return?" And I didn't respond at that point—it was like 5 in the morning—I was like, but my response was like, "David, I did a design. I painted a picture. Like, if you painted a picture of a car, it can't drive, right? You know that." And, like, that's the point that he—and it's funny because I just said it. I literally just said this to Paul B: just because David says something doesn't mean we do it. And that's true. Just because David says that something should be done doesn't mean that it is done. Likewise, just because David is expecting something doesn't mean he always gets it for various reasons. We try sometimes, but like the challenge is he doesn't understand the cost of building stuff and how long it takes him. He doesn't both understand or delve into the weeds enough to understand that. Like, oh, so you're working on these filters. You're not also working on the agent?

No. Like, that's...

Well, because the filters—once you search, you'll do other stuff with the agent. Yeah. Like, just because I bought a house doesn't mean I have kids now. Like, there's a process we think, right? So just keep that in mind. It's not a—with him especially, it's not a like he just gets ahead of himself and ends up kind of communicating in a little bit of a clunky way, is all I'm saying. So, you know, keep that in mind.

Yeah, I hear you. That is also hard. So that's, I'm glad to see that coming out now.

Yeah. Yeah. Do me a favor. Your honest take on the Salesforce stuff because I'm still a little confused. Not confused. I'm a little like—why like some of these problems seem like we're just doing the same problems over and over again. And I'm trying to understand that. So, you know, your thoughts there would be helpful.

So there's that. What else?

Kind of. And I'll tell you about it. You want to hear about it?

Okay. Okay. Thank you. Give me a sec, yeah. I don't think there's a ticket for it. Okay. So I am working on something. Something and I'll show you what your next piece of work is. Hey, it's this. Not the front end. It's related to this. And actually I'm going to switch screens now. I was just trying to be cute. It's this. This is what I'm working on right now: I literally took the designs and I listed all the processes that we were kind of showing. And then I was like, "What else could we add?" And I even asked Claude, "Do you have any ideas about..." And Claude came up with some interesting ones that were like, "Look at the PR—like the LinkedIn posts of people to see if they're champions or skeptics of our company." And I was like, "That's crazy." Which is funny.

Oh, by the way, yesterday in the swarm call, they told us they had a PE fund when they were trying to sell the PE fund as a customer. And the PE fund was like, "Can you, if we have you follow all of our employees or something like that on Instagram, can you look in their photos and identify potential people they know?" And David literally was staring at him and goes, "Well, that's creepy." I was like, "Yeah, that's creepy." PE companies never proving to anybody that they're not the worst people in the world.

Yeah. So, anyhow, I think this is a duo project. I'll work on the front end, you work on the back end. This goes back to what we were talking about a couple weeks ago about like, "Hey, what do you want to work on? You want to work on data-related stuff, right?" Well, some of this is—well, most of this is data, but it's also orchestration. And so what I want to do, and I just pulled this list here. I haven't gone through it with David. I'm going to run through it with him. And then on Monday, I want to sit with you and say, "Okay, let's put an architecture—which you've already done in some ways—for this: the atomic unit of a process that the agent can run or stream or whatever you call it." And I want to sit down and be like, the inputs and outputs are always structured in this way. It's always going to be a worker in Cloudflare or it's going to be an async worker or like something like that. It's going to have these types of tests. Like I want to make this—I want you to be a brain-dead, cookie-cutter programmer by the time you're done with this list, so to speak. Not because like, I want you to be brain-dead or whatever, but because we want to make it like a machine. We can just keep adding processes as we add more data potential partners and stuff like that. Also because eventually we will hopefully be able to hire somebody else and they would be a junior for you that could like literally be the guy—I know all you've wanted in life is to just, you know, manage down. I can tell.

No, I'm kidding.

But the idea is, so David was talking about the swarm yesterday, and he asked—I was like, yeah, they just have so many integrations. I'm like, this is how you do that, David, because this is what we do at Webby. They have a Polish CTO who hired a bunch of Polish engineers in Poland into her cheapest shit, though not as cheap as they once were. And you literally go, hey, Michal, like, you're the clay guy, and hey, Brennan, you're the gong guy, and that's all they do, you know? Or maybe they do two or three, but it's like kind of how we have time working on the Chrome extension or the Salesforce thing. That's how you would do that. And so it's easier at scale when you have a little bit of money because you can just hire really cheap people to do it. So in theory, eventually we'll build up the team like that.

But the goal is, if I start building the interface over here, we need to fill it up with stuff. Most of this stuff—I'm just going to share my whole screen because I keep switching back and forth. Most of my note-taker I'll just hide that. Most of this stuff is what should show up in theory here. And it's not like a list of all 20 of them, but when certain things are running, it would lead to the react, you know? For instance, being able to allow somebody to reply and cue an intro right now—I can build that from here. I mean, you could build it too, but I can build it, design it, do all that stuff. I could do this too, but you've been doing this and you're better at this stuff. You think like an agent does, so this is better suited for you, I think. I can do the interface, I can do all this stuff, but if we can collaborate on this and really frame it out—what's the input, what's the output, where's the data stored, how's it handled, et cetera—then I think we add this to a page, and it's the one item running for a week, and then the next item next week it's three items, and the next week it's ten items. You know, we just add them as we go.

Does that—is that work you're interested in doing?

Okay, good. Yeah, so the whole point, and this is a little different than we originally talked about. To be frank, I don't know where the Rally interface is going. I think the interface eventually just moves over to Slack completely. And it's still on the website, but it's really Slack. These guide teams are where the orchestrator just is always running. And so we'd still show the orchestrations over there too, but it's like different use cases, I think. We'd have the orchestrator running and they turn it on for a deal team. So maybe, and this goes back to the use case and the pricing they've been working on, maybe you could just run it on a connector and that's free. But if you're doing it for a deal team, it's all of these 20 things, or 10 or 15 or whatever it is. And they're running 10 constantly, but the point is it's not going 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15. It's 1, 2, 3, 4, 5. These are the 15 things that the agent can do, and based on the information the agent has, it enqueues those processes as it sees fit. So, yeah, it's definitely very orchestrator-heavy, but it's also not like just cron-job everything. It's how do we build the supervisory agent that makes reasonable judgments? Like, hey, we haven't run the map key personnel job in a while. We should run that, you know? Or we just ran the map key personnel process, and that takes three bucks per company to run. So therefore don't run it anytime soon, you know?

Yeah. I think so. And you know, we're still bickering on words, because David found the deal team. And then in a conversation we had, he was like, yeah, so you have your internal deal team and then you have ours. And I was like, we can't call it what they already call it. We can't do that. And he's like, well, no, they'll know it's different. Like, they won't know it's different. We've worked with clients for years now who get confused by the fact that we try to use what you call a lead is not what anybody else in sales calls a lead. So let's stop using the same word. He's like, well, it's like a relationship deal to him. Like, that's better. And now he's saying relationship map, which is probably the better answer because that's not an overloaded CRM term. So, you know, I don't know that that's the right name for it, but it's part of it. Maybe it should just be Target Company.

Yeah, I agree. It's weird, and I'll be honest—I don't mean it as a slight against David, but I don't mean it in an insulting way—but David isn't a CRM-first kind of salesperson, and running a company that builds software for people who are CRM-first people is a hurdle. You really have to know the domain you work in.

I don't know how much David knows about CRMs. He doesn't totally know how CRMs work—that's what I've learned.

Yeah, yeah, I don't know. As we get bigger, we hire a product person too that will help and guide it. I think we'll see.

Cool. So that's kind of what's going on next, right? And hopefully that will be both interesting and exciting. And it's also something that we can collaborate on together so that we can support getting it done. And not only getting it done, but getting it done iteratively, which is something we've struggled with at times, right? It's like really making sure that we're working in small chunks.

Yeah, we'll be fine.

Cool. Okay, I actually have a lot to chat about today, to be completely honest. Anything on your mind you want to chat about?

Okay, that's good. I got to look into the security stuff. I appreciate you pushing out yesterday. I ran it, and oddly enough, I don't know if it worked or not. It sent it to you.

Oh.

That's rude. We just set up your peer review bot. Gotcha. Yeah, I'll do that. That's fine.

What I need to figure out, and I honestly just don't know much about it, is what's going on in this tab here. Like, obviously we have the dependent pod alerts for vulnerabilities, but I'm pretty sure we want the Breakman stuff coming in here too, because this isn't a code quality thing, but I don't know if it's supposed to show up here too or what. But it is what it is. I'm glad that's running. And if it's not messaging you, we need to get it to message the DevCon channel or something, but that's good.

Cool. And then what else? PR-wise, I'm going to push up that other one for the fantasy rollout for people re-indexing. And then the other thing is I'm going to try this weekend to find some time to look at the analytics cache process because it's so slow. We have to do it for every person, not thousands of people, but the fact that each takes a few minutes is a problem.

The connector index is a problem because it's doing it for everybody, but if you just do it on the account level, that's the same data for everybody. So that's an easy fix. But the reason I want to look at it is it suggests there might be improvements in the database, like indexing or extraneous data or something. So I'm going to look at that. That's really it on my plate.

The other thing is they haven't responded—I gotta check. I'm pretty pissed about this Replicate thing where they're billing me. Did I tell you? I've mentioned it. I had it canceled, but I'm going to keep it canceled if they're going to bill us extra. They want to bill us an extra $29 because I have 28 flex reviews.

You're hitting that limit. It's your fault.

Because you're—it's not your fault, but like, you are the one doing stuff. But it's on my PRs. I'm the one being slightly more PR prolific, but you're the one who's acting on the PRs more prolifically.

Yeah, I'm who missed. We'll see. We'll talk to you later. Okay. Big news if you need me.

The SmallWorld team held their daily standup followed by a 1:1 between Cameron and Kim. Kim completed the ReliAi search view filters and is merging the RallyEye filters PR, while Todd is finalizing Salesforce 2.8 cleanup and consolidating configs. Cameron recapped the Swarm CEO/CTO meeting — the team has no interest in joining Swarm but sees potential in a revenue-sharing data partnership. The bulk of the conversation centered on a new "agent orchestration" initiative: Cameron is defining a list of background agent processes (distinct from the user-facing concierge) and plans a duo project where he builds the frontend while Kim architects and builds the standardized backend workers. The SOC 2 audit issue was resolved, the demo environment was killed, and there's an ongoing Greptile billing dispute over flex review overages.