Why Every Traded Personal Agents for One Company Agent

Subscribe to Every (free for 14 days): every.to/podcast-every-offer?utm_source=podcast&utm_medium=audio&utm_campaign=companyagentepisode On this week's episode of The Every Podcast, Dan Shipper sits down with Willie Williams, Every's head of platform and one of the driving forces behind the Every Agent, which launched this week. They discuss how Every went from everyone running their own OpenClaw to relying on one shared company agent—and where they think personal agents belong in everyday life. Links to resources mentioned in the episode: Willie Williams on X: https://x.com/bigwilliestyle The Every Agent: https://every.to/agent?utm_source=podcast&utm_campaign=podcast&utm_content=1006companyagent OpenClaw: https://openclaw.ai Tend, Dan's open-source feed app: https://every.to/tend?utm_source=podcast&utm_campaign=podcast&utm_content=1006companyagent Every’s Vibe Check on Dots: https://every.to/vibe-check/vibe-check-dots-always-on-agents-in-chatgpt?utm_source=podcast&utm_campaign=podcast&utm_content=1006companyagent Dot from OpenAI: https://openai.com/index/introducing-dots/ 0:00 Introduction 5:58 How OpenClaw spread through Every 11:13 Why everyone's personal agents eventually died 14:51 The difference between building for builders and building for operators 16:20 Why work will run on company agents while personal agents will stay personal 19:56 How engineering management has changed in the age of AI 21:43 Using an AI-powered feed to see everything happening across a product launch 23:53 Why Willie's token usage fell off a cliff 29:05 The benefit of having separate AI coaches for each part of your life 31:45 How Every is building its agent to surface information unprompted

• 0:00Speaker 00

  The hot thing right now is Dots. They're persistent agents from OpenAI. I was there at Dev Day when they launched them. I interviewed Sam about them and I've been using Dots since before they came out and now we're into week two of Dots. So how's it going? If you read my vibe check, you know that I thought they were a little bit rough to begin with, but that the idea of a persistent proactive agent was obviously the future. Even before Dots launched, they'd become my main way of interacting with ChatGPT and they're still that way after launch. I'm really excited about the little things they've been doing to make the experience less buggy and more complete and I think it'll get even better

• 0:41Speaker 00

  over the next couple of weeks. What's been really interesting though is seeing the team at Every. We've got about 30 people, seeing them start to adopt it and some of the use cases that have started to emerge. So in case you're looking for some use cases, I actually had my dot collect some use cases to talk to you about. So one thing that Dots are really useful for is keeping up with school emails. So Kieran Klassen, who works at Every, says that it scans all of the school emails he gets from his kid's school and tells him what he needs to know. Another thing that Dots are really good for is catching access requests or like little things, little messages you might have missed in Slack and making sure that you get them. So Ariel Shipper, our head of operations,

• 1:23Speaker 00

  her dot caught a few people asking for different accounts across the company and asked if she wanted it to just approve access. Dots are also really good for working when you're not at your computer or your hands are full. So our head of video, Randy Kunsman, was running around yesterday trying to get our launch video out for the Every agent which launched and he was doing work for that launch just on the phone with his dot, talking it through different video edits and different things he needed to get done. I also think, this is something that came up in my interview with Sam and I feel this too, Dots are really good for filtering out all of the noise that you don't really want to see and only showing you stuff that really matters. So for

• 2:04Speaker 00

  example, I was doing a lot of model testing over the weekend and normally model testing requires me to be in the Every Slack, which, once I'm in Slack, it's just like this rabbit hole of I'm just going down doing tons and tons and tons of work. And what was amazing about using my dot doing model testing was I was able to post the results of all the experiments I was running in Slack and if someone responded, that my dot would tell me about the response, which would help me stay up to date on what the team was thinking without getting me sucked into the whole rabbit hole of everything going on in Slack and that helped keep my weekend pretty clear and calm. And I think that's a huge, huge thing for Dots. So what's the overall vibe? If you don't have a dot, I highly recommend you get one, especially if you're in the ChatGPT

• 2:48Speaker 00

  ecosystem. If you're using Muse or Instinct or Grokbots, you should check it out. I think that they're pretty much at feature parity right now, but there's something really good about having something that's hooked into the entire rest of the ecosystem that you use for your work and your life. There are still rough edges though. The whole permission structure of Dots is like completely a mystery to me and is extremely annoying. So for example, I often have to give multiple approvals to my dot to do something basic and even when I give it approval, it says it can't do it. So they really need to figure that stuff out to make the experience awesome. But I would expect in a year, this is the default way that you're interacting with all your AIs. You're probably not going to have it be a cute character. That would be my other like

• 3:31Speaker 00

  little hot take. Mine is named Boo. He's super cute. But my experience using OpenClaws and other named agents like this is that as soon as something better comes along, you throw away the cute little character you developed. So my guess is I'll be using a persistent agent in a year, but Boo will probably be dead. Rip Boo. We've also actually been working on our own persistent agent. That's not a personal agent, it's a company agent and that's actually come from this whole journey we've been on, starting with OpenClaws inside of Every and everybody having a personal agent and realizing how valuable it was for the company to have a personal agent. So we built one for ourselves and then we just released it to all of our subscribers. So I brought Willie Williams on, our head of platform, who was instrumental in

• 4:13Speaker 00

  bringing the agent to life to talk about personal agents versus company agents, what it's like to build an agent at the frontier right now and what that means for engineering teams going forward. I hope you enjoy.

• 4:38Speaker 00

  Willie, welcome to the show.

• 4:40Speaker 01

  Thank you for having me.

• 4:41Speaker 00

  So for people who don't know, you are the head of platform. You are one of the driving forces behind the Every agent, which we, well, by the time this comes out, we'll have launched. It's launching tomorrow. And I want to talk to you about the whole agent phenomenon because I think we're, we're in this world right now where personal agents are the hot thing.

• 5:03Speaker 00

  And there's like a little bit of Groundhog's Day for me, I assume for you, because we went through the, our personal agent phase like nine months ago or eight months ago with the original OpenClaw phase and I feel like we learned a lot from that. And then now I feel like there's this whole new era of people who are experiencing the same thing and maybe will learn the same lessons. And we, for us, we went from OpenClaw, everyone in the company's got a personal agent to we have a company agent and we liked the company agent so much that we actually just made a company agent for other people to use. But I want to just talk about that and try to talk about it as honestly as we can because it's possible we're wrong. But I do think we've

• 5:43Speaker 00

  learned some things that are, that are important. Maybe let's start with, with OpenClaw because we actually disagreed and I think you ended up being right. So tell us about like the OpenClaw situation and sort of your take and then, and talk about why I was wrong.

• 5:58Speaker 01

  Well, I think that the funniest thing was actually at the time is called Claudius? Claudus? Claudus. It was like the earlier version before, because it was Clawbot, right?

• 6:10Speaker 00

  Oh.

• 6:11Speaker 01

  Right? Even before that.

• 6:12Speaker 00

  Interesting.

• 6:13Speaker 01

  It's called.

• 6:13Speaker 00

  And Moltbot.

• 6:15Speaker 01

  There was, we forget about the.

• 6:17Speaker 00

  Yeah, the two days of.

• 6:19Speaker 01

  Yeah. Big shout out to Ash actually, who was like, who I think first messaged me about it. And he was like, Ash is one of our contributing writers and he was like, dude, this thing is awesome. And I remember because I, I knew like Peter's name from like the top token burners leaderboard, like he was up there and just like, I was like, oh man, this is all, this, this folks are burning a lot of tokens, right? And then he was like, oh yeah, they're making, he's making this thing called Claudus. And I was like, oh, this is kind of a weird thing, you know, kind of follow the project. And then Ash came back and was like, oh yeah, dude, this is really cool. And so we started trying and then, and then I think Brandon was the first one was like, okay, I've got one.

• 7:04Speaker 00

  Yeah.

• 7:04Speaker 01

  You know, I've got one. I'm starting to like wire it up, right?

• 7:08Speaker 00

  R.I.P. Zosha.

• 7:10Speaker 01

  R.I.P. Zosha. And I think, you know, and then it really kicked off in January because we went to, went to Think Week, right? And for folks who don't know, it's like our semi annual gathering of Every where we sort of like think about the future, you know? And that was the place where we were finally all in person. We're like, oh, I can show you what's going on, right? And I think there's a lot, this is a recurring theme of like showing people how to use AI is like a very transformative thing, but telling them is sort of like lands flat. And so we had a bunch of people who were OpenClaw pilled who were like, oh yeah, this is, this is cool. Like this is what I do. I order my groceries. I've got it in my group chat with my, my family, you know?

• 7:50Speaker 00

  It really spread. Like we, we were all in Panama together and I, I remember coming out of that. I think there was a model drop on the day we left and the claws just started working way better. And then I made this like internal place for them to like talk and they started talking to each other in our Slack and it was like, whoa, this is crazy. Yeah.

• 8:11Speaker 01

  Yeah, there was, there was, it was, it was like a funny thing where we were like, okay, the first, the first blocker is it's just a little bit too technical. And so we went through a couple iterations of like, how do we do it for the team or the less technical members of the team? And there was like, you know, we're, we've got a template, so we're spinning up info for that. And then we made a little service where anyone could like enter, like kind of walk through people through the setup and then we get there. But we also like looking at it and we're like, oh man, security is hard, you know, for all of these things. But you know, we sort of like, I think we like powered through a lot of that. And I mean, I have this vivid memory.

• 8:51Speaker 01

  Of just like so many bots in our Slack, like so many bots, like absolute chaos. Cause, cause you also have more than one, Brendan being like, just, just having so many bots running around doing various things.

• 9:05Speaker 01

  And I think, I mean, I think the, the big takeaway was one, it was, it was so valuable for people to see other people using AI. And, and you know, up until this point, it was like, oh, I built this like gigantic like Claude or Codex Command Center on my local machine. I've got it doing everything.

• 9:24Speaker 00

  Yeah.

• 9:25Speaker 01

  But like, the only thing that people would see was the output of it. And so, once we had OpenClaws inside, I was like, oh, I didn't know I could use it that way. Oh, I, oh, that's actually pretty cool.

• 9:35Speaker 00

  And I didn't know that you prompted in this way.

• 9:38Speaker 01

  Yeah.

• 9:38Speaker 00

  Like, there's all these little things people do when they're using their AI that you never got to see. And it was like, oh, oh, I'm like learning every time I look in Slack.

• 9:48Speaker 01

  Yeah. And the, and the other thing I think that was, that was really big was, uh, you know, I think we liked our personal bots. And I think this is another thing that I think we're seeing now again, is that you sort of form, you're like, oh, I'm a, I'm attached. Well, mine was last, right? I'm like attached to last, right? And, um, and what it does, and it's like a customization. So for folks who are like tinkers, folks who are builders who like playing with these things, it was like a great, these setups are great, you know? For folks who aren't, folks who are like a little bit more in terms of like operator thing, it was just like, oh, this is, there's a bunch of security worries. It's super complex. How do I get it set up?

• 10:28Speaker 00

  It keeps breaking, like.

• 10:30Speaker 01

  And then if you don't invest in it, it like kind of, kind of fades. It like doesn't, it, it, it doesn't keep going with the rest of it. And I think that was, that for us was the striking point where it was sort of like, we want, instead of everyone investing in their own agents and making them better, you know? Day to day over time, it's like, why don't we take all that energy and put it into one agent that the whole company benefits from?

• 10:52Speaker 00

  If you're enjoying this show, you should subscribe to Every. Every is the only subscription you need to stay at the edge of AI. And for a limited time, we're giving all of the viewers and listeners of this show 14 days for free to get all the education and equipment you need to stay at the edge. If that sounds like you, you should subscribe. The link's in the description. I think also what happened is everyone had open claws. And for sure, like one of the coolest things was your claw became a little bit of an extension of you. Like it sort of, I remember one of the first people who, well, Brandon was like a big driving force here. And then the second person who like really took it to a new level was Austin, um, who was at the time our head of growth.

• 11:35Speaker 00

  And he had, he had like a growth claw that you could ask all growth questions to. And that was like really interesting. But I think, at least my memory of what happened is then we started, we were like, oh, we're gonna build a platform for anyone to make their own. Their own claws, even if you're non-technical, called plus one, R I P plus one. But I don't think at that point it was really a company agent. It was just like make a claw, like one click claw maker. And we tried that and we were having a lot of internal debates about, should it be one agent for everyone or an agent per person? You were very one agent for everyone. I was very like agent per person.

• 12:14Speaker 00

  I think you had a lot of hesitancies about the idea of building up our own platform infrastructure to support this like claw as a service thing. And we were like, fuck it, like, let's do it anyway.

• 12:26Speaker 01

  Yeah.

• 12:26Speaker 00

  Okay.

• 12:30Speaker 00

  Which I appreciate, but also like in retrospect was like not the best idea. And like we, so we launched that in beta and I think basically found that it was, there was a lot of demand, especially at that time, there was a lot of demand for it. But OpenClaws itself, it was very hard to build a platform on. And it was like moving so fast and it was completely vibe coded and like, it's just, it's just hard to make the entire technical stack. And then have some, have like an open source product that's not really designed to be a platform, like as the thing you're serving. That was like messy and through that process, I think all of our claws basically died.

• 13:11Speaker 01

  Yeah.

• 13:12Speaker 00

  Because we just stopped, I, I remember my claw's name R two C two. And I, I felt like, oh yeah, I had this like really attached relationship to R two C two. And then at some point, like someone knocked my, knocked the Mac, uh, the Mac mini power off. And then like it disappeared and I, and I, but I didn't restart it, you know? Cause like at that point I was using Codex and I was like, this is fine. Like I don't need that.

• 13:37Speaker 00

  And that, that was like a really interesting thing for me to realize like, oh, um, okay, there's all these people in the org who don't really, um, who don't really want to maintain their own claws. And then the people who do want to do it, like eventually you kind of get tired of it if there's a better option. And what we ended up doing is we, I think we were always starting to use like Claude or Codex. And then there, there are a couple of these company agents that we tried that were like really sort of changed how we thought about this. So Victor being one, which was just really, really well designed.

• 14:11Speaker 00

  And I think at that point, we started to realize, oh, like having a single company agent makes a huge difference.

• 14:18Speaker 01

  Yeah, there's a lot, I mean, it even, even just boils down to sort of like naming, right? Remembering which agent had which expertise was, was.

• 14:27Speaker 00

  Annoying.

• 14:28Speaker 01

  Was annoying and hard. And I remember like Austin's, Austin's agent, Montane, right? Was like, oh, it sort of became the, the, the gravitational center because he built it up to be like so good at the growth job. That it was like, well, I could add, I'm already using it for myself, like minor growth jobs.

• 14:48Speaker 00

  Yeah.

• 14:48Speaker 01

  Like I could invest in my agent or I could just add on to Montane and everyone else.

• 14:53Speaker 00

  Yeah.

• 14:55Speaker 01

  Sort of gains from that. And I think, I think there were a lot of pressures there. But of course, like counteracting this, but again, was like, I think we really liked our personal agents. You know, we're like, we're, we're, we're trying to keep it there. But I mean, the other, the other thing I think is, is underappreciated is that, um, when you're building sort of like AI products now, the surface area for like products and agents is really large. You know, before it was like, oh, I've got a SaaS app. I can, I can sort of like map out all the user flows and you go and build it. And you sort of know how users will interact with your product. And the AI world is just, and with plus one, what we noticed was, oh, we didn't even realize that you wanted to go do this thing. And like, how do you support it?

• 15:33Speaker 01

  So there was always, I talked a lot about like complexity, like building and managing complexity for builders versus like operators is a, uh, is a hard job. Like builders like want to be exposed to the complexity. So because that's everything you want to do is going in there and tinkering and turning knobs and whatnot. But if you're trying to run a business, you're, you're not, you're not really, you don't want to do that. You know, mostly you just want something that just works for you. And I think over time we realized was for us, we wanted something that, that just worked. We, we didn't want to deal with all of the like little knobs and bells and whistles. And, and for the, the jobs that we were using an agent for, like a lot of product or project management, operations and whatnot.

• 16:17Speaker 01

  It was like better to have something solid.

• 16:19Speaker 00

  Yeah. And, and to have it be sort of shared. So when you use it, you're pretty confident that it's the right thing to be using for your company.

• 16:29Speaker 01

  Yeah.

• 16:29Speaker 00

  And someone has thought about it, that, you know, thought about the skill or the feature or whatever. And so now we're in this place where we're launching the every agent and saying that we use every day, like many, many times a day. The company would kind of like fall over. We didn't have it. And we have a bunch of beta customers. And we're also in this moment where, for example, I'm using my dot all the time from, from, uh, from OpenAI. And there's Muse and Instinct on the like personal agent side. One of the things, I'm curious, I'm curious for your bet here. One of the things that I feel is I think I will be using something like my dot in a year in Slack. Specifically because I use it inside of, like, it's basically just a portal to my Codex.

• 17:15Speaker 00

  But I probably won't, mine is named Boo. And like, it probably won't be named Boo. Like Boo is gonna be dead in a year. Like lifespan, six months, maybe, you know, cause I, the personal thing or the personality thing is cool. But particularly in a work context, I mostly just care like, does it do the thing?

• 17:32Speaker 01

  Yeah.

• 17:33Speaker 00

  Um, and the minute there's like, like, I'm like, oh, I have this personal connection. But the minute there's something that performs better, if it doesn't, even if it doesn't have it, I'll just switch. I think there's much more room for that on the truly personal side.

• 17:47Speaker 00

  So I think that's my bet with, with thoughts is like, you still actually need a company agent, like the every agent. And then that needs to play well with like everyone having their own like dot or dot like thing. That is basically a portal to their command center. What, what do you think?

• 18:04Speaker 01

  Yeah, I think, um,

• 18:07Speaker 01

  I think we can see the line between like professional and personal sort of shaping up very solidly. I think on the professional side, I do think it's sort of like a one agent per like company or kind of structure because there are so many advantages again to like it learning the workflows and what you do and everyone contributing just by working with it and making it better that it shapes itself to your company. And it's like the advantage of bringing a personal agent in, it's just gonna be, it doesn't have that context here. It's gonna be hard to teach. Like it, it needs to have been there a year to really get up to speed. And so I think we're gonna have that for the workplace. And I really don't see like

• 18:50Speaker 01

  us or the industry repeating the chaos of like every worker gets their coworker and we're all gonna like double our company size, you know? But on the personal side, I actually think the, I was talking about those, the forces, the opposing force, the one of personalization, I think is, is actually huge. I feel like the thing that you love is that you have an agent that is just yours. And you know, before the downsides there were like, it was open call, how do I do security? How do I, how do I manage it? How do I like run it? You know, there are all these questions that, that prevented it from gaining larger adoption. And now that you have bigger players coming in and more money, it's just like, oh, they've, we're starting to get closer to like a better consumer personal agent experience. And it's

• 19:35Speaker 01

  gonna be incredibly attractive.

• 19:37Speaker 00

  Yeah.

• 19:38Speaker 01

  And so I think, I think we're gonna see a split. I think we're gonna see like, I go to work and I work with my company agent at the company because we all are working together towards this collective goal. And at home, it's, it might just be me or me and my family that is using a personal agent to help us do very similar things, but it's like personalized tasks.

• 19:55Speaker 00

  One thing that I'm really interested, we've been talking about this a lot. I think you have a very good perspective on, and I just want to hear more about it, is you've had a long current tech.

• 20:08Speaker 00

  Doing engineering management at like great startups, bigger companies, you've founded your own companies, you led the engineering for this effort. With a team of, I don't know, five to seven ish people. How has engineering management changed, especially over the last couple of months? And what have you learned in the process of bringing this to market you didn't necessarily know?

• 20:32Speaker 01

  Ironically, engineering management

• 20:35Speaker 01

  feels the most similar out of, out of any of the jobs that I've done before in the present day, especially like now getting to come back to it and see what it's like in the age of AI. I think, I think there are lots of advantages, like

• 20:52Speaker 01

  for an engineering manager, a lot of your, your leverage comes from the ability of how much of the field you can see. And so in previous years, you would spend a lot of time trying to gather that information.

• 21:02Speaker 00

  What would be an example of something that's on the field that you need to see?

• 21:05Speaker 01

  Oh, you know, there are a lot of Slack channels, you know, like keeping up with Slack channels and emails and, you know, customer reports and how do you, how are you doing the integration of all of this into a single place? And like, like I run a 10 feed, right? That, that does this where it consolidates, it consolidates that all into a single feed. And during the day, I can just kind of go through and it really keeps me, gives me a much better like vision into everything going on related to the project. It's like, okay, is, is marketing happening over here? Like, are we, are we seeing customer reports over there? Which normally would have been, you know, meetings, right?

• 21:42Speaker 00

  That makes sense. Yeah. So, so for, for people listening who don't know, Tend is this like open source app I made that you can create basically custom feeds for yourself where you hook up data sources. So in this case, it's like meetings or, you know, Notion notes or anything like that. And then your AI can just automatically turn those data sources into a feed that like either has decisions you need to make or is like FYI, this thing happened in a meeting, here's, here's what you want. And it sort of learns over time what you want. So you can make a feed for yourself. So you have one for your engineering stuff. I have one for just like the full company. What are a couple of things that you've learned, maybe even in the last week from like reading that you might not have known otherwise?

• 22:19Speaker 01

  Yeah, a lot on the like marketing side of launching a product. So like normally when you're, when you're building a product, you're sort of in it with your, your teammates and you're like, you're looking at screens and designs and user error bug, bug reports and whatnot. And, and you kind of have a feel for everything going on. But launching a product, there's, there's a lot of things and conversations and meetings that happen that are public, but they're just out in places where you, you don't necessarily have the time to check or, or the attention. And so it's like, how do you go and grab those things and just make sure that like, one, that, you know, that they're going the way you want them to. And two, that you know who to give guidance to. Like sometimes, you know, the rumor mill of a company like comes back, it's like, oh, did you hear we're doing this sort of thing?

• 23:03Speaker 01

  And it's like, well, no, we shouldn't be doing that. But then who do I talk to is like a huge, a huge question. And as a manager, that's your leverage, right? It's like, I, I need to be able to go and talk to and set guidance and set vision for all this. And so having sort of this, this higher like field marshal level view of a lot of things is just so helpful, even, even for a small team.

• 23:26Speaker 00

  Yeah. I think you've been the beneficiary or I don't know if beneficiary, maybe it's the opposite beneficiary on the receiving end of a few times where I'm like, hey, I like read this thing in Tend, like that this happened in a meeting, like what's going on here? Or like, just, just to put my thumb on it or whatever. It's, I mean, for, as a manager, it's like, it's like the coolest thing. And it can be like weird for employees. But like, yeah, I'm very curious to see how that changes. Now would be a time as a manager and engineering manager where you could actually like be ripping tokens while you're managing a fairly large team. Like what's, what is your take on that?

• 24:03Speaker 01

  I think a lot of it comes down to people, you know? I think the, the job of the manager is still like a force multiplier on folks. And so it's like, where do I spend my like epsilon extra piece of energy? And the, it's, it's really probably not on tokens for coding, right? And, and it's surprised, this is the part of this actually surprisingly very much the same is going, it's like a people, still very much a people job. And you want it to be a people job. Now, the difference is your, your leverage has increased because their leverage has increased. And so the most valuable things for, for me recently have just been like, what are the bottlenecks? Right? How do, and how do we go and remove the bottlenecks? And so it's funny as I sort of transitioned from like doing more technical work back into

• 24:48Speaker 01

  engineering management, like you see my token use just fall off a cliff, right? Because I'm, I'm, I'm using it and it's sort of gathering all this information. But when it comes to action, like there's no, there's no billion dollar, well, sorry, billion token move I can make. It's, it's mostly like, we should, we still need to sit down and like have a conversation, you know? Now, everyone, it can like, there's a whole bunch of routing functions in there that are now, you know, help with AI. Like we can do summaries, like how do we, how do we summarize and pass information? All the, all, a lot of jobs that managers used to do and a lot of the things that used to be responsible for in terms of like ops and

• 25:28Speaker 01

  and the health of the team in the company like are, are much less manual now.

• 25:33Speaker 00

  Have you automated that?

• 25:35Speaker 01

  I mean, for the, mostly for the, for the agent team, it's just like, we're, we're good. Like I, it was just amazing to me having built a lot of dashboards and a lot of learning systems. It was like, oh no, we're like, because it's really definable and really testable and what agents are great at.

• 25:50Speaker 00

  Like what, what's an example of, because I don't, I, to be honest, I don't look at a lot of like agent team health metrics that you're looking at. Like what's an example of some of those dashboards?

• 25:59Speaker 01

  Well, some of them are, are what you might call like performance dashboards, right? Like how many errors are, are API throwing, right? Is like a, okay.

• 26:10Speaker 01

  Like there's a dashboard for that. Same on the product side. It's like, oh, how many users are using them? But the real magic now is before it was like you had to set like, like alerting systems were, were even, even just Fuzio were still pretty like neolithic, right? It was, it was sort of like, I'm gonna set an alert, I'm gonna set some bars, you know, it's usually above this or below this. And one place that just the magic is, is there is just like, you can put an agent, like look at this every five minutes and see if there's an anomaly. If there's anomaly, do intelligent routing. They'll like alert me, you know, like there are always things like setting up like on-call structures where you're like, okay, this person needs to be on-call and this person needs to be

• 26:52Speaker 01

  on-call. And now you have a, like you have an agent on-call as the first person all the time. You don't have to worry about vacation. You're not scheduling around. All these like sort of below the line pieces of engineering management have gotten much better.

• 27:06Speaker 00

  That's interesting.

• 27:07Speaker 01

  I want to throw it back at you. I'm curious, we were sort of on opposite sides before. Where, how do you think the, this personal agent dynamic is gonna play out?

• 27:14Speaker 00

  I mean, my take on technology is you're never wrong. You're just too early usually. So I'm still holding, holding out that I might be right. No, I think my, I think my current feeling is a big, and I, and I, I think you probably disagree with this, but a big reason why we went from everyone has their own to a company is they're just too hard to maintain.

• 27:40Speaker 01

  Yeah.

• 27:42Speaker 00

  And that we will start to go from, there's a company one, which will continue to exist. And then that will start to subdivide into like teams, which we already kind of have. Like we sort of also have a couple of these agents like Claudie or Bill, or there's a couple of other ones that are floating around that different teams maintain that are just like for that team. And then that will probably again bifurcate a little bit and, and maybe that's, maybe that's just an, a dot, an extension of your, your personal codex, but maybe it's something else. So I do think we will probably get to the shadow org chart at some point. I thought it might be a more bottoms up, like everyone has their own and then we kind

• 28:23Speaker 00

  of like coalesce into, but now it seems like actually it's more likely to be, you make something really great that's at the top and then you kind of split that out every once in a while when you, when you need the specialization.

• 28:34Speaker 01

  Yeah. I mean, there's sort of details to a little bit of what we've seen of like folks who are trying to do like AI transformation, right? It's like, usually if, if you're a line engineer, you're not the one trying to do AI transformation in your company. You're like, usually it's the CEO or someone high up is like, I, we really need to accelerate and like, like we're struggling to do it. And like, let's, let's drop an agent in here that like everyone can use and see, see other people using. I'm also

• 29:05Speaker 01

  Really intrigued by like if people are going to have like separate agents for each role. So like one of the things is funny enough in my, on the personal side, like in Grokbot, I use Grokbot. And I was like, I will run like, I have like a lot of coaches, right? And for whatever reason, that mental model is like, works really well for me. Maybe because it maps to humans. So it's like, you have a, you have a coach for this, a coach for this.

• 29:30Speaker 00

  What are your different coaches?

• 29:33Speaker 01

  I'm like, I'm a terrible swimmer, which is really funny because I like to surf, but like,

• 29:39Speaker 01

  I'm like a really bad lap swimmer for a reason. And I was like, okay, I need, I need someone who's going to keep me on like, like a lap swimming cadence.

• 29:47Speaker 00

  And why, but like, why not have that just be like part of a bigger bot that like does a bunch of different?

• 29:52Speaker 01

  Yeah, I mean, this is, this is where it's like an interesting sort of like, why is it psychologically that I'm like, I need this sort of like this separation for this one role? You know, it's like, do I have that? And then I have like a therapist, it's like, do they talk together, you know? And it's funny because I can kind of line them up on jobs to be done, but like, not in the way that I think that like, if I rationally was trying to explain it, what it would look like. You know, I have to deal with a storage unit. I don't have like a storage unit bot that's dealing with it. But like, I do for, you know, maybe it's like, is this a throwaway person? Do I, do I expect to never like interact with them again? If it's transactional, it's like a thread in, in Codex.

• 30:32Speaker 01

  If it's like, I'm building a relationship, you know, one day I will, I will FaceTime my, my Grokbot into my workouts, you know?

• 30:40Speaker 00

  That is interesting. Yeah, I think that there's something about, it's almost saying to yourself, planting a flag, this is an area I care about and will care about for a long time, that maybe you want a little bot for. Right now, the, my, my closest thing to that is like pinned tabs in Codex. But now those are totally outmoded because I just use dot. But there's no like, you can't have multiple dots. So.

• 31:03Speaker 01

  And I also wonder how much like context comes with it. So it's, I think the really interesting thing that's happening now is we're, we're ratcheting up the amount of productivity we allow. And so instinct and Muse and Grokfire are doing these things where there's like, they're starting to reach out.

• 31:18Speaker 00

  Yeah.

• 31:19Speaker 01

  And I think when you have AIs are reaching out, if it's all in Codex or Claude and it's, and it's coming at you. And it's like, you have to remember what thread it was. But if it's like, Willie's swimming coach is reaching out, like, I'm like, okay, this is a swimming coach. You know, I can instantly place you in the conversation of like, okay, I didn't, I didn't swim enough, you know?

• 31:44Speaker 00

  Yeah, I love, the productivity thing is a big deal. And that's something that we, we are just baking into the agent that I'm super excited about. Cause and this is very related to tend to, like, there's this whole thing that's starting to happen, which is, it used to be that in order to make a feed of anything, you had to be a gigantic corporation with like hundreds of millions of users and a lot of money. And now everyone can have their own feed and companies like ours can, you know, for example, in the every agent, it has frontier alerts that are just like, okay, like this thing happened, here's what it means for you, which I'm super excited about. I think there's like so many places to go in that sort of proactively telling you what you need to know at the, at the right time.

• 32:26Speaker 00

  That it used to be the AI was not able to do that in a way that was good. But now with Jeb and some of these other things like decisions API, it's like happening.

• 32:34Speaker 01

  Yeah, I mean, yeah, I mean, we very much have like a feed based world, right? Across everything, uh, which is, and it's a tension because, you know, I spend a lot of time on X, like probably everyone, I assume.

• 32:49Speaker 01

  And for me, it's like, it's a funny sort of explore exploit problem, which is like, like I'm on there mostly to like, see the new ideas emerging and hear how people debate them. But there's also a lot of trash and like, you know, like my like cognitive defenses say that like, they have to be up. Like I have to have my protections up when I'm reading it, just because it's like, I don't want to be dragged into, to this, you know, flame war or this, whatever. I'm, I'm trying to sift out the nuggets from this, this stream of draughts. And it's like, it would be great if I didn't have to be that, that thing, right? If I had, if I had an agent that could take the brunt of that and then I get out the nuggets.

• 33:27Speaker 00

  That's actually something that's been helpful for me with, with my dot is like, it does, especially, I mean, this is not a feed, but like for Slack, like just telling me, here's a couple of things that you probably missed.

• 33:37Speaker 01

  I think Slack's a feed too.

• 33:38Speaker 00

  So helpful, it's so helpful. Yeah. Willie, always a pleasure.

• 33:43Speaker 01

  Always a pleasure.

• 33:44Speaker 00

  Thanks for joining.