He Drove to Strangers’ Houses to Install AI Agents

I have crazy ADHD and I just kind of zone out all the time and my brain's just operating. You're the founder of Setup Claw. Setupclaw deploys OpenClaw for SMBs. The Hermes agent is going viral.

What do you think is next? The next bigger change is that right now when we think of agents, we think of these personal agents on your Mac Mini or your personal computer. Longterm, most of these agents will be in the cloud. So the question is how do we make it so a person can scale up from three four five agents to something like what composio has internally.

Composio internally has you know dozens of agents in the cloud. Let's go back to your point of the expansion of agents beyond small business owners and beyond. What I'm noticing now is that out of the many many SMBs that exist in the US, most of their problems right now are solvable with AIS, but there is way too much friction for them to solve. And what does this mean for founders building or who want to build in this space?

There's going to be an order of magnitude more agents. These agents are going to be much more autonomous. They will be they'll be shaping this product because they are the users, not the humans. Okay, Michael, I'm going to do your intro a little bit differently.

I'm going to start the sentence and you finish it. Okay. You ready? Okay.

You're the founder of uh Setup Claw. Setupclaw deploys OpenClaw for SMBs. You've deployed dozens of OpenClaw agents for dozens of SMBs. And the integration layer that makes it all worth it.

It's Composeio. Yes. Sweet. Okay, let's start with this.

Yeah. You've deployed a a dozens of open claws in people's um on people's computers, the cloud. Yes. And did you take the page out of the Collison Brothers playbook?

Oh, absolutely. Absolutely. I mean, I'm from San Francisco. Uh the Collisonson brothers are legends.

Everybody knows the story of how they came in in coffee shops and installed Stripe on people's laptops for their projects. This is exactly how we started. Originally our tagline on our website was actually call installs for OpenClaw. So the first few installations were in person.

So I'd go to your house, I would install OpenC cloth. And actually like side note, things were terrible until I used Composia, but that's neither here or there. So I would go install composio uh sorry I would install open claw I would install openclaw for people and it was very coliscoded so I would just go in figure out what they need figure out all the integrations a little later it became a little bit more standardized and I figured out okay how do we like set up security in a way that's good how do we make sure that you know the Mac is secure how do we make sure that we can deploy an agent quickly and how do we make sure they have all their integrations that they need to be successful and then eventually We scaled out once you do a lot of calls and installs. What's important is that you learn about your user.

So we learned, hey, what's important here is not just installing openclaw, but also figuring out what agents the company needs. So we started building out agents for individual companies. Yeah. We're in green apple books right now.

Yeah. Can you describe what this place means to you? I mean, it means a lot. It's the bookstore where I grew up, right?

So when I was a kid, basically I would come here three, four times a week sometimes. looking around trying to find a book. I would grab a free book downstairs in the bin. Maybe I'd go over there and read for a few hours and then maybe buy the book, maybe not.

I'd usually get a fourth way through before I made a decision just because I didn't have that much money to spend on books. So, yeah, I still come here a lot as an adult. You know, whenever I'm like hanging out with friends in the area, I like still try to pop in. These days, I don't really buy many books though.

Okay. More so like reading online. Oh my god. It's just like too much time spend I spend reading Twitter articles.

People need to stop posting engineering articles on Twitter because I get so much joy out of reading them. Oh my god. This is how we deployed AI agents to harness uh inside or outside the sandbox. Come on.

Well, I'll read that every day of the week. That's that's the stuff that gets me going these days. Wow. And do you remember what kind of books you would read when you were younger?

Oh my god. It was everything. It was like fantasy fiction. And at some point it was young adults.

And then there was like a lot of uh some philosophy. There was a lot of Nam Chomsky which is actually where my Twitter handle comes from. Michael Chomsky. Okay.

It was politics. All kinds of stuff. So there was like history from the Granny Smith area over there. There was, you know, fantasy over there downstairs.

Like usually anything in the front that was political. I really liked. That's awesome. Yeah.

Let's go find one of your one of your books that Let's do it. Let's let's try to walk around and see a book that I've read at some point. We'll we'll find one relatively quickly, probably. So, let's see.

This is a good one. Thinking fast and slow by Daniel Conman. Okay. Yeah.

He won the Nobel Prize for the idea that sometimes we lock in and think really hard and sometimes we're just like our brains are just running in the background. Now, for me, most of the time my brain is running in the background. So, I have crazy ADHD and I just kind of zone out all the time and my brain is just operating and I like might be eating something and then not notice that I was eating it. So, yeah.

So, this is this really is important to me because it taught me that hey, my brain is doing stuff all the time even when I'm not conscious of it. Okay. Um Yeah. And I think this is it's super important to just like do stuff and not overthink a lot.

This shows that uh yeah, you want to balance the two. So great book. I should read Yeah, I should read it. Should read the summary of summary.

I'll read the summary. Yeah. Sweet. Okay.

So before set ofclaw, what was the problem that you were running into? I think the problem I was running into is I was kind of dabbling with OpenClaw, but more so a lot of the people I was working around were dabbling with OpenClaw and everybody was having a bad time. It was very early on. The hype had never been higher, but in practice, people were trying it and having a terrible time.

So, I could see the potential was really high, but in practice, people were just not confident about the security implications. They would try it for a bit, but hit annoyances. Token costs were really high. They spent hours setting it up.

So, I was like, this this can't be a thing. So, I'd go in and I I thought in the beginning that, oh, maybe some people would uh take me up on it and I would do some calls and installs and then I'd have some fun doing it, but it completely blew up. My calendar was completely booked and it became very clear that this was an actual tangible problem to solve. Wow.

And let's talk about Ren My Header. Yeah. Where did that idea come from? Right.

My header is this incredible project. And the idea was from uh a friend Shish. And Shish is amazing at Twitter. He is incredible.

He's 10 times better than I am. He he gets retweeted by Elon Musk all the time. I It's crazy. U but his idea was like, "Hey, I have this Twitter header and it's valuable, right?

Because people are going to my profile. " And so he said, "Michael, I don't want to like be a developer. " And I was like, "Okay, great. I love this.

I'm a developer. " So I built this app for him. And ever since then, we realized that there were many problems with this. There's the creator marketing space is a very sophisticated place on Twitter where everybody's kind of paying to play.

Not a lot of people are doing like high quality organic content like you're doing right here. And a lot of people are just like paying for posts. A lot of people are, you know, trying to figure out, hey, if is there any kind of shortcut? Uh, an example is Perplexity, right?

Perplexity got kind of a lot of flack for just paying for a lot of creators. So, we kind of pivot to that. We try to like be the platform to like pay and organize these creators and do do that in a kind of high trust way for both sides. But, it's still it's still a crazy place.

So that business is like we have like thousands of creators on but our our revenue isn't great. So it's a very fun experiment. I have lots of hope for it. Um but it's still something I'm trying to figure out very much.

And how many things did you try between setup claw and re my header? I mean many things. many things. I have uh you know some software that's like I'm living off of um that is just like on its own.

Uh that was the one successful thing out of many and then I must have tried like dozens of things and most of them didn't work and even uh setup clause still very early. It's still still very much a proof of concept. We don't know where open claw is going to be in 3 months. We don't know how businesses will be incorporating it on a large scale and we're kind of at the forefront of figuring this out.

Do you think the hype with OpenClaw is going to like die? It has died a little bit. It has died and like the the best way to tell whether there's the level of OpenClaw hype is how many meetings I'm booking, right? So if like the number of meetings I'm booking is lower, okay, people are, you know, a little less hyped about OpenClaw, right?

So the hype has definitely died down. It's still there. I think now people are like, "Okay, this is cool, but what actual business problems can it solve for me? " Yeah.

Take me back to the very first set of clock client. Oh my god. The first client ever was Kelly. And Kelly was fantastic.

Kelly basically said, "Hey, come to this cafe. I have my Mac Mini. I need OpenClaw installed. And I I came with a friend uh because I was like, I don't want to go at this alone.

And together we spent 4 hours battling trying to set up everything uh for the first time. And I was pretty confident coming in. I was like, "Okay, I've done it for myself. " But no, it was actually a really big challenge.

And that was the thing that got us thinking okay we know after this us and everybody else is doing this poorly right now we did this and we didn't even understand the security implications of the time fully right and after Kelly we decided a few things one is well I decided a few things um we can't be doing this this way ever again not even for one more client I'm not going to spend going no no no going is fine but specific specifically um doing the installation in that way is unsustainable. So we had to figure out okay how do we make it a a lot faster and since then we started doing a lot of remote installations where we like go and like give them a tail scale link and do a lot of remote installations that was a lot more sustainable but more than that we needed to make sure we had you know some processes to make sure that the Mac Mini is secure that we could do this in a reasonable amount of time and that we had tools uh to set up integrations in a fast and uh suitable way. Okay. And so integrations like what is your integrations are very important because open claw is a tool is a a personal assistant.

That's it number one use case. It's used for business automations used for a million different things. But it's number one use case at the moment is personal assistant. A personal assistant is only as powerful as the tools you give it.

Right? And you need to be able to manage these tools independently of me the person who installed OpenCloud originally. So I need you to have some kind of dashboard to manage your integrations without me being present and like to add something. This this proved extremely important and also like simple things like setting up Slack.

You know we can't be doing this for 3 hours every day, right? Every time. So the question is like what do we need to do to make this 10 times faster so that like Kelly was really patient with us and you know the friend's car got towed right after that. So the money went completely down the drain.

We covered the some of that. Uh but uh you know we knew future customers wouldn't be as patient because we also learned that we needed to charge way more for what we're doing. So that was the other thing. So after Kelly we immediately doubled prices and then I doubled prices again and then I doubled prices again.

At some point we had an odd number or something and not an odd number but a weird number for our prices and I was and I was asked why is the number so weird and I had to explain hey every time we had too many meetings I would just tell cloud code to double the price. Wow. And so Kelly was non-technical right or Kelly was not really technical. Um I think she was somewhat technical but you know not super technical.

What does she What's the thing they always like what did um after setup did Kelly or any of your other clients say I didn't know it could do that like what do they do they say something like that this is extremely common so out of our clients my favorite type of client is I call this a super user and these are the people that we like set them up we created some agents for them we get them started and then a week later they're messaging us hey I have like five agents doing all these things across my one or multiple businesses and these people they know their specific domain in their lives 100 times better than we do. So the things they're doing are very creative like during our intro call like we talk about one or two or three things they do those two or three things and they move on to more ambitious things. Uh so it's very impressive to me how much non-technical h uh people can achieve with open claw completely non-technical zero coding ability. What's a one example of something someone achieved?

Oh my god. We have this one client who works in healthcare and they had this like compliance use case which was they they talked early on they were like hey I don't know whether I can solve this problem or not but if it could even like slightly help it would be massive. Now they have a dashboard with OpenClaw managing this internal compliance dashboard saving them multiple employees dozens of hours per week potentially saving them you know thousands of dollars per month if not like tens of thousands. This was it was crazy how fast they were able to achieve this because they were able to achieve this mostly independently right and in the matter of a few weeks with you know some initial help from us but really it was mostly them and this is not something you see in the AI automation consulting space where the client is doing the work usually it's the contractor doing the work and trying to figure out what the client needs it's it's kind of crazy because it flips this whole AI consulting automation thing on its head where you're not trying to do things for the customer, you're trying to enable them to do it themselves.

Okay. And so you mentioned how after day one you said you're never going to do this again with something to connect all their integrations. Yes. How long did it take you to find Composeio?

And what did you try before it? Uh Composio was the first thing I tried because uh for Composio I was already working with a team to try to find them at Devril. So uh the de I found them didn't work out unfortunately. Uh but yeah I was already working with the team.

I was already aware of the product. I knew what it did. Uh and in the beginning it was a bit of a not a great time. It was good.

It was it massively saved us a lot of time but it wasn't great. Uh because before Compose when I started it was kind of just for businesses and products trying to create their own integrations. you guys didn't have this consumer proumer side of like oh add integrations to your own open clock which ended up being extremely important for us uh later on when you added it I remember I had a meeting at some point and during this meeting I noticed that the composio UI had changed because I was onboarding like one or two customers a day to compose you manually um over like zoom or something and I noticed that the UI had changed and then they can now you know self onboard right and have their own like custom. We were doing some hacky things with API keys that was honestly not very ergonomic.

Uh and I remember doing it. I was like, "Okay, this is this is going to save me like 20 minutes per meeting. " So, yeah, it was is very cool. And that's when I kind of decided, okay, Compose is definitely the the direction to go cuz you guys were moving in the right direction.

We already had you in our Slack. You guys were responding to feedback and it felt like you guys were on our team. Awesome. And is it right that you were essentially misusing Composia for a while bending a B2B product to work?

Yeah, it's 100% true. We were completely misusing Composio. So Composio, as I said in the beginning, was completely this like separate product port. You have an app.

Let's make it so you could set up integrations. We were we were ignoring that and saying, hey, you have an open claw, right? We need some way to set up integrations in some reasonable manner. We know GOG CLI sucks.

uh we know all these other integrations are terrible, right? We don't want to be clanking out, you know, vibe coding integrations every time. So, what we did is we like got them an API key. We set up our own like, you know, uh little plugin into OpenCloud to make Composio work and it worked.

It was fine, but Jesus, it was so hacky, right? It was so hacky. Now, OpenCloud users can benefit of being able to do this without someone like us kind of uh guiding them through the process. Okay.

So, you mentioned how so like when Composio's UI changed, it it changed when you were mid onboarding. I was mid onboarding. Midboarding. Mid onboard.

And then I I went the new direction. I was like, okay, whatever they have has to be better than what we're doing and it ended up being a trip. Yeah. No.

So, now we use the new one and we use, you know, the original Composio for our apps. So, we have two apps. So we have our openclaw cloud app uh startclaw and then we also have our new app which is moving off of openclaw to give us a little bit more flexibility uh and deploying agents for some enterprises that don't necessarily need all of openclaw's features and need more reliability. Okay.

And in your own words, you probably already said something like this, but what does composia do for setup claw? How would you explain it to someone who doesn't know how to write a line of code? Your agents need tools. These tools need authentication.

Someone needs to manage these authentications. You need a middle layer. You need a proxy. You need some someone responsible for this.

Um, in traditional authentication, this would be something like work OS, right? You have this like middle layer that handles and owns your authentication. With agents, it's even more important because if you take these API keys, these credentials, and you put them on the actual computer where the agent is stored, this is a massive security vulnerability. But it's also an ergonomic vulnerability for your users, right?

So you need Composeio to serve as this middle layer. And the same middle layer is valuable if you're building a product, right? So instead of writing integration code for every single integration, you can just use Composeio, let them handle that. Let them make sure agents can use these tools effectively.

Okay. Yeah. The Hermes agent is going viral. What do you think is next?

I think what's next is pretty simple. It's two things. One, the agent ecosystem is going to be less dog Right now, use OpenClaw, you have a bad time sometimes. Use her Hermes agent, you have a bad time sometimes.

" It's a completely a software issue. Like it could be 10 times better, but the capabilities of models have grown so much faster than we can keep up with software, which sounds crazy, but it's true. So things will kind of just work in the future. They'll be more reliable.

People will have a good time and stuff like Hermes or OpenClaw will be less of a big decision because both will be less bad, right? And then there'll be like other players that are also good. And then the next bigger change is that right now when we think of agents, we think of these personal agents on your Mac Mini or your personal computer. Longterm, most of these agents will be in the cloud.

So the question is how do we make it so a person can scale up from you know three four five agents which even you know top cos have only five agents right now to something like what composio has internally. Composio internally has you know dozens of agents in the cloud. We we need to get to the point where companies have, you know, 100 plus agents, a thousand plus agents in the cloud working actively where you can not worry about being bankrupt from tokens where they're providing value in some way both in this is going to start first in software companies because they're already doing this internally. Most you know thoroughly advanced software companies already have this in some form like composio especially agent first companies.

These are mostly like SF, New York companies. Uh eventually this will expand and it'll expand past these very AI native companies to the point where nonInative companies and regular SMBs and enterprises will have many AI agents in the cloud. So when we get to the point where you know some SMB that is making like 10 million AR in a boring sector like HVAC and they have 50 agents doing something across SEO, marketing, content, uh managing people, you know, payments, whatever like that is where we're going, but we're very far away from that and that's kind of where we're going as our business. That's our bet.

We need to move towards that reality which it's we're going to move to that reality naturally because it's where agents are going and it's the only way agents can scale. can't scale on your personal computer, but it needs to go in this direction uh in a way where the developer ecosystem doesn't become scattered with a bunch of different agent SDKs that are like not good and there needs to be some kind of standardized way to you know figure out which sandbox vendor you use. Figure out where your file system is if you're using something like a fuse mount or a remote file system to like make this extensible. So we need to figure out how to make go back to the days of something like the open AI AI SDK where agent development is intuitive and accessible to everybody.

And then we need to figure out how do we orchestrate agents in a way that they're actually usable for regular companies. Right now we have really cool projects like paperclip um and even Hermes and open cloud people are pushing them very far. I would say these are in the proof of concept stage, not in the enterprise value stage. And there's a lot a lot a lot of engineering work to do before they get to that point.

Okay. And what does this mean for f founders building or who want to build in this space? I think what this means is you have to look a little bit longer term and at least like the one or two year uh period where there's going to be an order of magnitude more agents. These agents are going to be much more autonomous.

These agents will be making product decisions. They'll be figuring out, hey, like what software do we use? Do we use Composer? Do we use something else?

Well, the agents will decide, right? They'll be giving feedback. They will be they'll be shaping this product because they are the users, not the humans. Humans will have much more uh many more agents.

Individuals will have many more agents. They'll have agents for finance on their phones, on iMessage. They will have agents in the cloud. And uh developers will have many agents autonomously managing their like software in many ways looking after PRs, doing research, thinking out how code could be better, educating the developer on how they could be better as a developer.

This in order to scale exponentially, you have to be in the cloud, right? So, but you also need uh this connection between your personal devices and scaling up to the cloud so they're synced. So there's so many engineering problems and so many UX problems to be solved because the potential here is so much higher than we have what we have uh kind of realized so far. Okay.

Yeah. I think right now we're living in such a like big like in a bubble where only like it's I feel like it's not as mainstream right now. Would you say like it's not but eventually we'll expand past the bubble. So things start NSF and they expand out, right?

Even in this bubble, we're not pushing things very hard. We're not expanding past where we're not reaching the limits of what these current like generation of frontier alms are capable of. And we know the next generation of frontier is going to be much more powerful and are going to introduce their own challenges, right? So and opportunities.

So what we need to do is we kind of as a you know SF developer community where we have the power to kind of shape this future we have to think okay first of all we need to make sure that the software we're shipping is not you know vibe coded slop to like let people actually experience this magic without getting caught up in the trivial bugs right like hey like there's this absolute magic of open claw that people are missing just because there are certain things that don't work right we need to make sure that people find success and aren't, you know, blocked by these things. And we need to build things in a way where we're prepared for this next wave of more capable LMS. Let's go back to your point of the expansion of agents beyond small business owners and beyond SF. Yeah.

What I'm noticing now is that out of the many many SMBs that exist exist in the US, most of their problems right now are solvable with AI agents. They're completely solvable, but there's way too much friction for them to solve. For example, many of them just don't even have a decent website, even though that's kind of solvable with LMS at this point. when we go into the actual business processes, it is way too hard for them to in practice solve them with the current tools.

So, in order for this new wave of agents expanding to SMBs to work, it needs to be much more accessible to these SMBs, agencies, contractors, and everybody else to build these agents for these businesses. It needs to be an order of magnitude easier because right now it's far too hard, right? We're building out OpenClaw for uh businesses, agencies, etc. And it's the amount of work that we're doing to make this happen is it's quite a lot, right?

We have our own open call cloud, right? We have, you know, our own kind of ways we install things on uh for SMBs. And you know, if I'm a contractor and I just want to build some a solution for someone else and do it super easily, there is no great solution, right? Claw desktop is cloud co-work is great, but if I'm a contractor, this it's not a great tool for me to use, right?

If um there's not a lot of open source software that, you know, people can control well and modify, right? If I want to deploy something for an enterprise, same thing. There's going to be issues. So many people are making you know hundreds of thousands of dollars a month um much more than us.

People think of us as the open claw consultancy but we very much are not. There are people making much an order of magnitude more money than us in sectors like finance just solving these problems with openclaw. But one of the reasons we've kind of avoided this is that in many of these cases openclaw is just not the right tool for the job. It it can do the job.

It could do it well, but it's its form factor is just not right. It's not meant to be deployed in cloud environments. It's not meant to solve a lot of these issues. It reliability and observability aren't high enough by default, right?

And has all these things that are extra and not relevant. And openclaw is moving in a direction where this is more possible. Hermes agent is um possibly but still still not the right thing. If we want to move to a world where everybody has access to this, it needs to be way easier to ship and someone people really need to like organize and some standards and figure out okay what are these use cases and what are some software where we can make it really easy for people to solve those specific use cases with one standard framework protocol that is maintained by the open source community and agreed upon.

Right now there's so many different half-ass solutions and it's very frustrating for us. Like you can think of us as an agency and a software company at the same time. But as an agency, if we didn't have our own software, we we would have a terrible time. We would be so dissatisfied.

There's OpenClaw clouds that exist that just go to agencies and solve their OpenClaw hosting problems and like fix OpenClaw when it breaks. There's there's so many like band-aid solutions that people are building just because these things these solutions have struggled to caught up to the reality of these models and how fast they're moving. So if you're a business trying to like automate something, focus on your skills. Focus on the thing you're automating.

Focus on just making it work and don't really worry and overindex. Is it open CL? Is it Hermes? It doesn't matter.

figure out make it work somehow. Make it make sure you have a skill that automates the process. Whether you'll be using OpenCloud or Hermes in six months, I don't know. There might be something better.

I I I won't even pretend to know. Right. These these things will get much better. What are some examples of like products that people um that these people can build these?

I think the beautiful thing about these agent native companies is that the agents become customers and then if you're taking the bet that there'll be you know an order of magnitude or two orders of magnitude more agents and maybe I'm being naive maybe there'll be three to four orders of magnitude more agents than there are now then these products their growth potential is extremely high compared to traditional SAS and this is part of the reason some of these have massive valuations because people are making the bet that you know the number of agents will increase. If you can build a product like that, you will have an extremely strong place um in a world that is moving towards agents. Uh examples are Composio, Exa, Parallel, agent mail, stuff like that where okay like you're clearly building stuff that agents need but like we know this building stuff agent need agents need is not a new idea. If you can build stuff to make it easier for these regular SMBs or individuals that are away from tech to capitalize off these gains of, you know, agents improving, that's an even better bet in my opinion.

If you can make it so that an SMB has is able to uh deploy a marketing agent with very high reliability and just be vertical, it's fine, but be vertical and confident. Don't be vertical in general. and you can do one thing well and go after, you know, there's so many SMBs out there and just do this one thing for them with agents. This is a very viable business.

Is it a venture scale business? It may or may not be, but is it an extremely good business to have? Yes. What do you mean when you say vertical like be Yeah.

So, horizontal businesses are something like us, StarClaw, um these like agent hosting uh services, create an agent, lot claude, co-work, open work, whatever. uh and they're meant for everybody, right? And they don't solve one solution well, but they're making the bet of like agents are so capable. We don't know your specific problems.

You'll be able to solve what a specific problem you have on these general frameworks. And they are correct, but there's going to be a million of these general general frameworks, right? Some of them will have advantages, some of them will have disadvantages for a particular person or use case. If you have some kind of domain expertise, right?

Let's say you're a lawyer or whatever. Um, let's say you're a marketer. You could create a very vertical niche product for this exact problem, right? And then you could go and do outbound and you can scale that, right?

And sure, the total addressable market is smaller, but you're playing a very different game because you're out of the rat race. The rat race when you're building horizontal agents is everybody's improving at such a rapid rate. You have to be constantly pushing code and just trying to stay relevant. This is the situation with conductor T3 coat whatever.

Um these are all like and these are coding agents but there's also general horizontal agents like we're building one for example. If you could escape that rat race by creating something vertical for a specific use case that lasts and will be a problem in 2 years and you'll still be better than a model that is you know two times three times smarter than it is now. Then your life becomes much more easier. You have a clear ICP.

You have a clear problem that you're solving and your user doesn't start with a blank screen chat chat box that looks like everything else. Oh, what do you want to do today? If you could avoid this chat box in some way, you have a much higher chance of winning, right? You still want the chat box to appear because you want it to be customizable.

Like it's 2026, you need it somewhere. But if you could come in and provide immediate value for a customer instead of having the customer come in figuring out, okay, what they can actually use you for, that's going to be a big deal. And that's where companies are kind of missing things. They still depend on the user to provide value.

No. Come in and provide value immediately and let the user customize. What's an obvious about AI agents that no one is paying attention to? I'm trying to think really hard about this.

Let let me think. I mean, I think people are underestimating how little you have to do to have a successful product around AI agents. Uh, for example, Victor AI is super successful. It's literally Slack with a cloud agent.

Like, it's a very unsophisticated product. People are going to make a lot of money off uh a text message agent in some vertical finance, whatever. Like, you don't have to do everything. But the major thing that I think people are missing is how similar the base product is going to be.

It's going to look extremely similar. Historically, uh SAS was just a dashboard with Oth, right? It was a dashboard with O. It might it had all the same things, but you had off, you had a dashboard, you had some components, you have some charts, whatever.

in the f now it's looking like all these AI products are moving in the same direction where there is now the element of okay you need some kind of um sandbox enabled powerful agent maybe it has a sandbox maybe not but definitely has some kind of virtual file system uh it is you know much more powerful in terms of this hands or ability that it used to have it used to is just like an AI you know SDK simple like open AI SDK just like chatbot now this is isn't even enough for a base product right and now you have this idea okay like even products that are like very irrelevant to these like horizontal agent platforms will have a lot of the features of these horizontal agent platforms which is deploy an a schedule an agent for later do stuff like this um so the baseline of SLA SAS has kind of moved especially if you're building agents in some kind of vertical. So this means the amount of work you have to do to create you know a SAS in these days especially in some agent related niches is much larger. So the baseline of what the SAS is has changed and then people are only scratching the surface of what people you could actually achieve with you know generative UI and customized dashboards and stuff like that. So TLDDR I think what people are missing is that the baseline of what SAS is um the things that are now table stakes have changed right there's more you have to build just to be competitive so you have a strong take of what deil is I do I do your position if a devril person can get their own sponsors they're not doing devril they're doing field marketing explain okay okay this is my friend Nick's uh opinion I will say it's a very very strong opinion.

So the scope of Devril is so large. I would say what you're doing right now is Devrell. You would call yourself uh community what would you caller storyteller right this nice title I consider under it under deell but like it the title no longer has any meaning because it has expanded to so many different positions of so many people doing different things. There are people who call themselves deell that are doing like LinkedIn outbound like is that an SDR or is that a de right?

So it's just like the scope is is so large. I would say if you're a very good Devril, you can probably get your own sponsors. I I will say that's true. It's either you have your own YouTube channel, you have your own Twitter, you have some kind of social media where if you're really good at marketing and devrolls typically market themselves as a brand, you should be able to get a brand deal hypothetically if you wanted to.

Okay. You're also running a Devril job board, but it's just Yeah, just you. It's a Devrell job board. Uh I it's it's there.

I mean people have found deval jobs on it which is really cool. Okay. Uh I also like run a very unsuccessful Devril recruiting agency where I find people devs. I originally found one for Composio but the guy became too successful and couldn't get the job building an AI agent company.

I mean you can't make this up. Um so yeah it's it's a lot of fun. If like you're someone's passionate out there about Devrell and wants a job or uh if you're a founder looking for Devrell, contact me and I'll keep you in mind. It's very much I think of it as an arranged marriage.

It's an arranged marriage of and it's very hard because but once it's a right fit, it's a right fit. Is Devril a new ter a role and you I've never heard of devs before like is this a new thing? It's kind of a new thing. So it used to be PLG productled growth and then it was renamed Debrell.

Uh some people in the industry say that it was renamed in order to seem less intimidating and salesy. So it was Devril is somewhat of a salesy role like you're at the end of the day leads to sales relationships with developers. You're hoping at some point that leads to enterprise sales or developer sales or or something of that sort, right? So it used to be a PLG productled growth and then uh eventually it was renamed Devril to seem a little less salesy I would guess.

I'm not really sure but when I was entering it was already more called Devril than PLG. That said when I started two years ago my team was called PLG. It was a PLG team u but I was called a developer advocate and then the team was renamed Debrell like a year later. Mhm.

You're a regular at Bay Area AI meetups. I mean, I am somewhat regular. I'm like in San Francisco, I guess. I would say that's true.

What have you gotten from showing up in person that you genuinely can't get online? Not much. Not much. If you're out there going to AI meetups, I I question what what you're doing, right?

Uh and I say this as someone who's been completely desensitized to it. As someone from San Francisco, I've been seeing these for I don't know, my entire life. I don't see much value these days. is if I go to an AI meetup, it's like I I'll probably see some friends there, people who go to every single one, you know, who are who are good.

Uh, a lot of these AI meetups are completely waste of time. Uh, some of them are fun. Occasionally, you'd go to a fantastic meetup and you learn a lot. You learn stuff you can't get anywhere else.

Uh, there was one at YC that I went to a few months ago uh about RL that was just fantastic. Occasionally, you'd go to an event that is fantastic. I went to Milala. I had an incredible amount of fun.

Uh, but if you're going to just every single random event on a Tuesday, you're probably wasting your time. That's my opinion and people will disagree. And I've ran these events before. I've thrown these events.

I mean, if you want free pizza, go ahead. Have you ever met someone like very well known at any like meetups? Anyone like really cool that you look up to? I mean, yeah.

There people famous people at least in the Twitter sphere uh are at these events all the time. I think there was an openclaw event that I uh wanted to go at some point but wanted to go to but I had to work late and I later found out that what was that actor from Two and a Half Men? You know who I'm talking about? That actor from Two Two and a Half Men?

He's like super famous. Charlie Sheen. No, the other one. The the one that came in after the Charlie Sheen left and replaced Charlie Sheen.

Ashton Kutcher. Ashton Kutcher was at an open claw event, right? And I missed that event and I could have potentially met Ashton Kusher. So yeah, there's there's some cool people like CEOs of major tech companies.

Uh next JSC comp is always fun because it's like all of these like enterprise CEOs. It's a lot of fun. A lot of cool people connections you can make. Okay.

So building in public, you do it, right? Yeah, building public's great. What's the risk you accept when you share your process online? Have you ever shared something you later regretted?

Uh, have I shared something I later regretted? Uh, maybe I can't put my finger on a specific happened. So like, yeah, I can't put my finger on a single and I like what I have like thousands of tweets. I can't put my finger on a single one.

I kind of regret. Uh, I have deleted some, but like it wasn't something I really regret. Yeah, I think building building in public is probably fine. These days, people will copy your product really quickly.

I do kind of maybe regret maybe not putting setup claw and start claw on trustmr. What's that? Uh trustmr is a platform where you can verify your revenue and I think we got some traffic from it but I question its value. I very much question its value because the idea is that you show builders that you're getting revenue but in reality people go there to steal ideas, right?

So it's like you're just asking for competition. So we'd go there and then instantly someone like two people copied our exact product or three people copied our exact product and then you know maybe some some people would land on the wrong domain name. Um so in general I would kind of avoid that. I would post on Twitter.

Um I would be very careful about sharing revenue numbers. Okay. Uh but that you know build in public the Twitter you can get your first million dollars in revenue in Twitter. Absolutely.

Yeah. What's something new you're starting to keep building your name in the space? Definitely my podcast with my friend Nick. So very excited about that.

We're doing weekly episodes. Uh very much unproven. There is very little evidence that we'll succeed, but we'll do it anyway for as long as possible. Uh what else is there?

What do you talk about on the podcast? Like what's like the for the premise? Like what's the I don't know. There's it's a once a week tech podcast and there's always some news, right?

The format is just us talking. Usually we have some major topics of like X or Y events like we're both chronically on Twitter so we have some knowledge of these events and we have opinions on them. A lot of engineering stuff, a lot of agent stuff, a lot of things that we just actually think are really interesting like sales, deell marketing. Uh, I think it's probably really good for startup founders and people in the AI AI agent space because that's the things those are the things we're kind of like focused on as people in the space.

The first episode you did, you quote retweeted uh someone I forget his name, but he said that almost everyone's doing podcasts now. Like what's your take on like on that? And yeah, I mean a lot of people are doing a podcast. I will say podcasts are something they're a long game.

They're not something you find instant success in. So, we we're kind of here for the long game. We're only doing one episode a week, which may or may not be enough to succeed. Uh but the question is, can we do enough in this once a week podcast to be successful?

If not, we have a massive amount of fun doing it. So, uh it's it's a excuse to hang out with Nick, who's busy. We make time for each other. We make time to do it.

It's a lot of fun. That said, can we do better at it? Can we make a better podcast? Absolut lovely.

Like I I want to do 10 times better. Keep getting better. Keep getting better. Yeah.

Yeah. Yeah. But it's so much fun. Yeah.

We actually have Nick on the on the podcast uh tomorrow. We're filming with him. So, I'm excited to hear about his takes. Sweet.

Thanks so much, Michael, for joining this podcast in this very unusual location. Yeah, of course. Thanks for integrating me into your podcast.