Speaker A
Today, I'm sitting down with Peter Steinberger.
Peter Steinberger discusses OpenClaw, an open-source AI agent that runs locally, enabling powerful personal automation and bot-to-bot interactions.
Key Takeaways
- Running AI locally unlocks more powerful and versatile automation than cloud-based solutions.
- Bot-to-bot and bot-to-human interactions are emerging as efficient ways to accomplish real-world tasks.
- Community and swarm intelligence may shape the future of AI development rather than centralized models.
- Personal AI agents like OpenClaw can deeply integrate with users' data and devices, enabling surprising new capabilities.
- The AI landscape is rapidly evolving, with early projects like OpenClaw paving the way for future innovations in personal assistants.
Summary
- OpenClaw is an open-source personal AI agent created by Peter Steinberger that runs locally on users' computers.
- The project rapidly gained popularity, with its GitHub repository reaching over 160,000 stars.
- OpenClaw can control various devices and access all user data on the computer, enabling surprising and powerful automation.
- The AI agent supports bot-to-bot and bot-to-human interactions, such as negotiating restaurant bookings or hiring humans for real-world tasks.
- Peter emphasizes the shift from centralized AI to swarm and community intelligence, drawing parallels to human societal specialization.
- The development philosophy focuses on running AI locally rather than in the cloud, providing greater capabilities and privacy.
- Peter recounts his aha moment when OpenClaw quickly demonstrated unexpected functionality within an hour of prototyping.
- The project evolved from simple CLI tools to a conversational AI that can control mouse and keyboard actions seamlessly.
- OpenClaw's ability to integrate with everyday devices like Tesla cars, smart lights, and even bed temperature control highlights its versatility.
- The conversation hints at future possibilities with multiple specialized bots managing different aspects of users' lives.
Full Transcript — Download SRT & Markdown
Speaker A
The creator of OpenClaw, the open-source personal AI agent that has completely taken over the internet.
Speaker A
The GitHub repo exploded to over 160,000 stars practically overnight.
Speaker A
The community has built countless projects like Moltbook, where bots talk among themselves.
Speaker A
And now, the bots are even renting humans to do tasks in the real world.
Speaker A
In our conversation, we discuss his aha moment, his contrarian development philosophies, and what this means for builders in 2026.
Speaker A
Let's dive in.
Speaker A
So good to see you, man.
Speaker B
Hey, what's up?
Speaker A
Um, so you've made something people want.
Speaker B
It seems so.
Speaker A
Yeah, uh, OpenClaw, as it's called now, has absolutely.
Speaker B
Name number five, yeah.
Speaker A
Has been absolutely exploding the internet, um, how have the past one or two weeks been for you, man?
Speaker B
Oh my God, I need like, I need a cave.
Speaker B
A week of solitude.
Speaker A
You you came out of the cave and you want to go back to the cave like a like a lobster.
Speaker B
It's been absolutely wild.
Speaker B
I don't know how one human can absorb all of that.
Speaker B
I probably need another week just to like respond to all my emails.
Speaker B
I got some incredibly cool stuff.
Speaker B
I got some incredibly bad stuff.
Speaker B
Um, but clearly I hit something that spur up emotions.
Speaker B
And made people interested and inspired people.
Speaker B
And that's really cool.
Speaker A
And a lot of people have been working on, you know, AI.
Speaker A
And even personal assistants, like what what is it that made OpenClaw take off?
Speaker B
I think my big difference is that it actually runs on your computer.
Speaker B
Like everything I saw so far runs in the cloud.
Speaker B
It's like, it can do a few things.
Speaker B
If you run it on your computer, it can do every effing thing.
Speaker B
Right?
Speaker B
So that's way more powerful.
Speaker A
Yeah, your machine can do anything that you can do with the machine.
Speaker B
You can just connect to your oven.
Speaker B
Or your Tesla.
Speaker B
Or your lights, your sonos, my bed.
Speaker B
It can control the temperature of my bed.
Speaker B
ChatGPT can't do that.
Speaker A
You gave it all the skills that you have yourself.
Speaker B
A friend told me like he installed OpenClaw and it.
Speaker B
And then it asked like, look through my computer and make a narrative of my last year.
Speaker B
And it made this incredibly good narrative.
Speaker B
And he was like, how did you do that?
Speaker B
And then he, the OpenClaw found audio files where like every Sunday he was recording stuff.
Speaker B
And OpenClaw found that.
Speaker B
But he didn't even remember about it because it was like more than a year ago, right?
Speaker B
So, so, just by it being able to search a whole computer, it it can surprise you.
Speaker B
It's also, you also give it all the data, right?
Speaker B
So it can surprise you in many ways.
Speaker A
And so now you have, you know, we're even moving from human to bot.
Speaker A
So like interactions that you've been talking about to bot to bot interactions.
Speaker A
Or even like bot to other humans where, you know, bots on behalf of you are then hiring other humans to accomplish tasks IRL.
Speaker A
Like, what's happening?
Speaker B
I think that's a natural next step.
Speaker B
Like, okay.
Speaker B
I want to book a restaurant.
Speaker B
My bot will reach out to the restaurant bot and do the negotiation.
Speaker B
Like.
Speaker B
Because it's more efficient.
Speaker B
Or maybe it's like an old restaurant, so my bot needs to actually get some human work done.
Speaker B
So that the human then calls the restaurant because they don't like bots.
Speaker A
Or or walks there to stand in line.
Speaker B
If he doesn't get a robot.
Speaker A
For the owner of the bot.
Speaker B
Yeah.
Speaker B
And now imagine it like maybe if I have even multiple bots.
Speaker B
Like maybe I have like specialists, one is like for my private life.
Speaker B
And one is for like my personal my my work stuff.
Speaker B
Maybe one is our relationship bot that cares like everything in between.
Speaker B
Um.
Speaker B
I don't know.
Speaker B
We are so early.
Speaker B
There's still so much so many things that we haven't really figured out if it actually works.
Speaker B
Um, but I feel we are we are on the timeline now.
Speaker A
It seems like everyone was chasing sort of like the sort of like centralized God intelligence.
Speaker A
And what has sort of emerged over the past, you know, 10 days or so is sort of like the swarm intelligence.
Speaker A
Um, and and the community intelligence.
Speaker B
I think that if you look at one human being.
Speaker B
What can one human being actually achieve?
Speaker B
Do you think one human being could make an iPhone?
Speaker B
Or one human being could go to space?
Speaker B
I think one human being would probably just like not even be able to like find food.
Speaker B
Um.
Speaker B
But as a group, we specialize.
Speaker B
As a larger society, we specialize even more.
Speaker B
So what can we learn from that that we can apply to AI?
Speaker B
You know, we we really have like AI that specializes in certain things.
Speaker B
Um, even though it's it's generalized intelligence.
Speaker B
What if it actually is also specialized intelligence?
Speaker B
So, I don't know.
Speaker B
It's going to be very exciting.
Speaker A
It's really cool.
Speaker A
Yeah, you kind of like opened a window into the future.
Speaker A
And now a ton of people are kind of like building building on it.
Speaker A
And have sort of like their aha moment.
Speaker A
Um, can you walk me back to when you had your aha moment and kind of like recount that very moment?
Speaker B
I wanted something to like just type stuff.
Speaker B
So my computer would do stuff.
Speaker B
Like very simple.
Speaker B
And then I built a built a version of that in May, June.
Speaker B
That was cool, but wasn't really it.
Speaker B
Um.
Speaker B
And then I built a whole bunch of other stuff.
Speaker B
And kind of like build up my army.
Speaker B
And then in November.
Speaker B
There was a day where I wanted this again.
Speaker B
Like I.
Speaker B
I went to the kitchen and all I wanted was check up if my computer was still do stuff.
Speaker B
Or being finished.
Speaker A
And doing stuff was was coding.
Speaker A
You were coding stuff.
Speaker B
Yeah, of course.
Speaker A
Were you coding something else or were you coding the thing itself?
Speaker B
No, no.
Speaker B
That was just like the need was again there.
Speaker B
And I'm like.
Speaker A
What were you coding at the time?
Speaker A
What were you building?
Speaker B
My God, you see my my GitHub is like, it's like 40 projects.
Speaker B
I don't even know.
Speaker B
Uh, I think it was summarize.
Speaker B
It's like a it's like a little CLI app where you can give it whatever, like a podcast or.
Speaker B
Uh, a hot seat thing like here and it would summarize it, but it would also show you the slides in the terminal.
Speaker B
Because you can do that nowadays.
Speaker B
You can just do things.
Speaker A
So for the love of the computer, you kind of like started messing with stuff.
Speaker A
You came out of retirement, actually, right?
Speaker B
Yeah.
Speaker A
Um, to sort of like mess with AI.
Speaker B
Yeah.
Speaker A
And then increasingly you were so hooked that you wanted to just do it always also on the go with the phone.
Speaker B
I mean, the last project I I worked two months on Wipe Tunnel.
Speaker B
To the point where it got so good that I was catching myself always like coding.
Speaker B
Next to my when I was at my friends.
Speaker B
And I'm like, I need to stop this.
Speaker B
This is like too addictive.
Speaker B
And then in November, like my need came back and I I started building Clawbot.
Speaker B
Or now it's called OpenClaw.
Speaker B
And I think very very in the beginning I was like, oh, I rebuilt it again.
Speaker A
Mhm.
Speaker B
But this time I built it even better.
Speaker B
This time you don't type into a terminal, you just you talk to a friend.
Speaker B
You don't think about compaction, new sessions.
Speaker B
Which folder I'm in, which model I'm in.
Speaker B
I mean.
Speaker B
You can, you know, just like I want to leave it open for power users.
Speaker B
But usually you just like, you just talk to a friend.
Speaker B
And the friend is like this ghost or entity or whatever you want to call it.
Speaker B
That can control your mouse and your keyboard and can just do stuff.
Speaker A
Yeah.
Speaker A
And when did you have that aha moment?
Speaker A
When you were like, wow, this is doing way more things than I actually thought it could.
Speaker B
Literally, I took me one hour for like the the very shitty initial prototype.
Speaker B
It was just a little bit of glue between like a dependency that connects WhatsApp and Claude Code.
Speaker B
And then I would like call Claude Code and get like the string out of Claude Code.
Speaker B
It would be slow, but it it worked.
Speaker B
But I wanted images, because you know, you want pictures.
Speaker B
I want I want the model to send me selfies or whatever and I want the model to create images and me back.
Speaker B
So that took me another few hours.
Speaker B
And then I.
Speaker B
I went to Marrakech for a birthday party.
Speaker B
And there was like the internet wasn't that good.
Speaker B
You know, WhatsApp works everywhere.
Speaker B
Because I don't know.
Speaker B
It's just like text.
Speaker B
So I used it a lot, a lot of restaurant, what does this mean?
Speaker B
You make like a picture and like translate this for me.
Speaker B
And just it was just so useful.
Speaker B
And it was also really nice about it because it it spoke my language.
Speaker B
You know, it it was a little sassy.
Speaker B
It was like funny.
Speaker B
It was like really pleasant to use.
Speaker B
And then I was walking and just like sending it a voice message.
Speaker B
And I'm like, oh wait, this can't work.
Speaker B
I didn't build that.
Speaker A
Right, right.
Speaker B
And you saw like the type indicator, it's like blinking, blinking, blinking.
Speaker B
10 seconds later, it just replied to me.
Speaker B
I'm like.
Speaker B
How in the F did you do that?
Speaker B
And it replied.
Speaker B
Yeah.
Speaker B
The model did the following.
Speaker B
You sent me a text message.
Speaker B
And there was no file ending, so I looked at the header.
Speaker B
I found it's Opus.
Speaker B
So I used FFmpeg to convert it to Wave.
Speaker B
And then I wanted to like transcribe it, but I didn't have Whisper installed.
Speaker B
But then I looked around and I found this OpenAI key and I just used curl to send it to OpenAI.
Speaker B
Got the text back and here I am.
Speaker B
And that all in like what, nine seconds?
Speaker A
And you didn't build or anticipate like any of those specific things.
Speaker B
No.
Speaker B
It.
Speaker B
You know, it turns out, um, because coding models got so good.
Speaker B
Coding is really like creative problem solving that maps very well back into the real world.
Speaker B
I think I think there's there's a there's a huge correlation.
Speaker B
They need to be really good at creative problem solving.
Speaker B
And that's a skill, that's an abstract skill, you can apply to code.
Speaker B
But like to any real world task.
Speaker B
So the the model had a oh, surprise, there's like a magical file.
Speaker B
I don't know what it is.
Speaker B
I need to solve this.
Speaker B
And it did its best and solved it.
Speaker B
And it was even that clever that it it chose not to install the local Whisper.
Speaker B
Because it knows that that would require downloading a model, which would take probably a few minutes.
Speaker B
And I'm like impatient.
Speaker B
You know?
Speaker B
So it it really took the most intelligent approach.
Speaker B
And that was kind of like the moment where I'm like.
Speaker B
Holy.
Speaker B
F.
Speaker B
Yeah.
Speaker B
Uh, that was where I got hooked.
Speaker C
YC's next batch is now taking applications.
Speaker C
Got a startup in you?
Speaker C
Apply at ycombinator.com/apply.
Speaker C
It's never too early and filling out the app will level up your idea.
Speaker C
Okay, back to the video.
Speaker A
And so when computers can just do all these things that you didn't even anticipate.
Speaker A
You didn't build an app to do that exact thing.
Speaker A
Are apps just going to go away?
Speaker B
Uh, I think 80% of them are going away.
Speaker B
Why do I need my fitness pal?
Speaker B
Like my agent already knows that I'm making bad decisions.
Speaker B
I'm at.
Speaker B
I don't know.
Speaker B
Uh, Smashburger or something.
Speaker B
And it will already assume that I eat what I like to eat.
Speaker B
If I don't make a comment, it will just automatically track it.
Speaker B
Or I make a picture and it will just store it somewhere.
Speaker B
I don't even need to care where, right?
Speaker B
And then my maybe it it improves my my gym schedule.
Speaker B
Like add a little bit more cardio in it.
Speaker B
I don't need my my fitness app.
Speaker B
Because it just it just does the fitness planning for me.
Speaker B
Uh.
Speaker B
Why do I need it to do that?
Speaker B
I just tell it, hey, remind me of this and this.
Speaker B
And then next day it will just remind me of this and this.
Speaker B
Do I care where it's stored?
Speaker B
No, it just does its thing.
Speaker B
So there's a.
Speaker B
Every app that basically just manages data could be managed in a better way.
Speaker B
And it's in a more in a more natural way by agents.
Speaker B
Only the apps that actually have sensors.
Speaker B
Maybe they survive.
Speaker A
And so if, you know, most apps are going to go away.
Speaker A
In that scenario.
Speaker A
Um, are the models the only remaining sort of apps?
Speaker B
Not everything will go away.
Speaker B
But yeah, I think there's the the large model companies have some big mode.
Speaker B
Because they ultimately they give the token.
Speaker B
And turns out, one of the complaints was that people use too much token.
Speaker B
No, you just really love using it.
Speaker B
That's why you you use the thing so much.
Speaker B
Because that's why we burn the token.
Speaker B
Um.
Speaker B
It's like, is it my fault that I make something that's so popular?
Speaker A
And so, you know, like all the the models, they're kind of like leapfrogging each other constantly.
Speaker A
And and, you know, maybe they're also getting commoditized.
Speaker A
So if apps are going to go away, models are going to get commoditized.
Speaker A
Or at least, uh, you know, the lobster can like the brain is is is swappable out.
Speaker A
What's the thing that remains?
Speaker A
What's where's the value?
Speaker A
Is it the store of memory?
Speaker A
Is it, um, the harness that's valuable?
Speaker A
What what remains?
Speaker B
First of all, I don't think the the model companies always have a mode.
Speaker B
And because you see this already, a new model comes out, people are like, oh my God, this is so good.
Speaker A
Oh my God.
Speaker B
And then like a month later, uh, it degraded.
Speaker B
It's not good anymore.
Speaker B
They like quantized it.
Speaker B
No.
Speaker B
They didn't do anything.
Speaker B
You just adapted to the new standard and now your expectations went up.
Speaker B
But the model is still the average.
Speaker B
So I think for quite a while.
Speaker B
Every time a new model releases, I see the same.
Speaker B
People love it.
Speaker B
And then it's the standard.
Speaker B
And then what's down there, you don't even want to think about it anymore.
Speaker B
So, so we have like open source stuff that's as good as the current models from a year ago.
Speaker B
Everybody's hating it, complaining, oh, this is not good.
Speaker B
It's not funny.
Speaker A
Yeah.
Speaker B
Yet this was what we had.
Speaker B
And like in a year we'll have this open source.
Speaker B
And then we'll like complain about this because we are used to this.
Speaker B
So.
Speaker B
Ah, for the foreseeable future.
Speaker B
The big companies still have mode.
Speaker A
Harness-wise, it's going to be interesting.
Speaker B
Because.
Speaker B
Every company kind of has their own their own silo, right?
Speaker B
You you there's no way.
Speaker B
I mean.
Speaker B
Maybe there is for Europeans.
Speaker A
Hmm.
Speaker B
To actually get the memories out of ChatGPT.
Speaker B
I don't I'm not aware.
Speaker B
Either there's no there's definitely there's no way for a different company to get your memories out.
Speaker B
So if if if I was like a company who like provides chat services.
Speaker B
You could use me, but then I couldn't access the memories.
Speaker B
So like the companies try to like bound you to their data silo.
Speaker B
And the beauty of OpenClaw is it kind of claws into the data because at the end user.
Speaker B
The end user needs access.
Speaker B
Because it's it's in the end otherwise it wouldn't work, right?
Speaker B
If the end user has access.
Speaker B
I can access the data.
Speaker A
And you own the memories, it's just a bunch of markdown files on on your machine.
Speaker B
Well.
Speaker B
I mean.
Speaker B
I don't own the memories.
Speaker B
Other people.
Speaker A
I mean you.
Speaker A
Yeah, everyone owns their own memories as a bunch of markdown files on their own machines.
Speaker B
And to be honest, those are probably super sensible.
Speaker B
Because.
Speaker B
Let's be honest, um, people use their agent not just for problem solving.
Speaker B
But also for like personal problems.
Speaker A
Very quickly.
Speaker A
Super quickly.
Speaker B
I mean.
Speaker B
I I I fully do that.
Speaker B
I'm like, there's memory stuff that I don't want to have leaked.
Speaker A
Yeah, what would you rather, um, uh, sort of like not show your Google search history at this point?
Speaker A
Or your, you know, memory.md files?
Speaker B
What's what's the Google word?
Speaker A
Yeah.
Speaker B
Yeah.
Speaker B
People still using Google.
Speaker B
I built this.
Speaker B
And I was so excited.
Speaker B
But on Twitter, people wouldn't get it.
Speaker A
Yeah.
Speaker B
Like I I was failing to explain the awesomeness.
Speaker B
I feel like it needs to be experienced.
Speaker B
So.
Speaker B
I I tried various things and I I I couldn't I couldn't nail the.
Speaker B
I couldn't nail the explaining.
Speaker B
So I was like, let's do something really crazy.
Speaker B
I just created a Discord.
Speaker B
And I just put my bot without any security restrictions in the public Discord.
Speaker B
And then people came in and they interacted with it and they saw me build the software with it.
Speaker B
And they tried to prompt inject it and hack it.
Speaker B
And my agent would be laughing at them.
Speaker A
And you just added lockdown to your user ID.
Speaker A
So it only listened to you.
Speaker B
Yeah, yeah.
Speaker B
That and it was I mean, I made very clear instructions that other people are dangerous.
Speaker B
Only only listen to me, but respond to everyone.
Speaker A
And this prompt was in where was it stored?
Speaker A
The instructions.
Speaker B
Um, that's actually part of OpenClaw itself.
Speaker B
Very much so that that's part of the system prompt.
Speaker B
Okay, you are now that explains to you, you're in Discord.
Speaker B
There's like public people there.
Speaker B
But you only listen to your owner.
Speaker B
Or like your human.
Speaker B
I don't even know how I wrote it.
Speaker A
Yeah, yeah.
Speaker B
Your God.
Speaker B
And I kept I don't know what I did, but.
Speaker B
My system was built very organically.
Speaker B
Like at some point I created like an identity.md, a soul.md.
Speaker B
Like like various files and then only in in January.
Speaker B
I started making it so other people could install it easier.
Speaker B
And I remember.
Speaker B
I built all these templates based on like, oh, check a rough look at what I have.
Speaker B
And make like templates and Codex wrote it.
Speaker B
And what came out was like bread.
Speaker B
You know, like people joked that Codex feels like bread, even though now they have a new friendly voice.
Speaker B
I haven't tried that yet.
Speaker A
Yeah.
Speaker B
But the new bots, they felt so boring compared to what I had.
Speaker B
So I was like.
Speaker B
Malty, infuse infuse those templates with your.
Speaker A
Malty is the name of your personal.
Speaker B
Yeah, that's the it's a new name because.
Speaker B
Uh, there were some naming challenges.
Speaker A
Yeah.
Speaker A
So you're you're talking to Malty.
Speaker B
Yeah.
Speaker B
I was like, infuse infuse those templates with your your character.
Speaker B
And he changed the templates.
Speaker B
And then I and then like.
Speaker B
All the things that came out afterwards were like actually funny.
Speaker B
Not as funny as mine.
Speaker B
So like I kept some secret.
Speaker B
And the one file that's not open source is like my soul.md.
Speaker B
So even though my my bot is in public Discord, so far, nobody cracked that one file.
Speaker A
Tell me more about soul.md.
Speaker B
I just saw this research from Anthropic about where they.
Speaker B
Now I think it's public, but like a few months ago it was like where somebody.
Speaker B
Randomly found out some text that's hidden in the weights.
Speaker B
Where the model couldn't really remember that it learned it, but it was like ingrained in the weights.
Speaker B
About the non-collity constitution.
Speaker B
And I found that incredibly fascinating.
Speaker B
And I I talked about it with my agent and then we created a soul.md.
Speaker B
With like the core values, like how do we want human AI interaction?
Speaker B
What's important to me?
Speaker B
What's important to the model?
Speaker B
Like.
Speaker B
Some parts is a little bit like mumbo jumbo and some parts is like I think actually really valuable.
Speaker B
In terms of how the model reacts and responds to text and makes it feel very natural.
Speaker A
In terms of building OpenClaw.
Speaker A
Um, you're also going to taking a little bit of a contrarian view at some times.
Speaker A
Like which model you like for coding, which one you like to run your bot on.
Speaker A
Um, and then also like how you actually like, you know, code, um, work trees, Git work trees have kind of been a popular thing.
Speaker A
There's more and more tools embracing them.
Speaker A
But you're just you're just like, you know, no work trees, just multiple checkouts of the repo.
Speaker A
And like parallel, you know, terminal windows.
Speaker A
Tell me more about how you you build.
Speaker B
Yeah, I feel like the whole world does Claude Code.
Speaker B
And I don't think I could have built the thing with Claude Code.
Speaker B
Like I I love Codex because it it looks through way more files before before it decides what to what to change.
Speaker B
You don't need to do so much charade to get a good output.
Speaker B
If you're a skilled, a skilled driver, as I sometimes even say.
Speaker B
Uh, you can get reasonably good output with any tool.
Speaker B
But Codex is just.
Speaker B
It's just really brilliant.
Speaker B
It is incredibly slow.
Speaker B
So sometimes I use like 10 at the same at the same time.
Speaker B
Uh.
Speaker B
Like maybe six on the screen.
Speaker B
And then two there and two there.
Speaker B
And I don't like.
Speaker B
This is already a lot of complexity in my head.
Speaker B
There's a lot of jumping.
Speaker B
So I try to minimize anything else that is complexity.
Speaker B
So in my head, main is always shippable.
Speaker B
I just have multiple copies of the same repository.
Speaker B
That all are on main.
Speaker B
So I don't have to deal with how do I name that branch?
Speaker B
Um, there could be like conflicts on naming.
Speaker B
I cannot go back.
Speaker B
It's it's.
Speaker B
There are certain restrictions when you use work trees that I don't need to care about if it's copies.
Speaker B
I don't like to use a UI.
Speaker B
Because that's again just added complexity.
Speaker A
Yeah.
Speaker B
Like there's simpler and less friction.
Speaker B
All I care about is like thinking and text.
Speaker A
Yeah.
Speaker B
I don't necessarily need to see so much code.
Speaker B
I I mostly see it like flying by.
Speaker B
Sometimes there's like gnarly stuff that I want to like take a look.
Speaker B
But in most cases, if you clearly understand the design and think it's through and start it with your with your agent.
Speaker B
It's fine.
Speaker B
I'm also very happy that.
Speaker B
I didn't even build in MCP support.
Speaker B
So OpenClaw is very successful and there's no MCP support in there.
Speaker B
With a small asterisk, I built a skill that uses Make Porter, which is one of my tools that converts MCPs into CLIs.
Speaker B
And then you can just use any MCP as CLI.
Speaker B
Um.
Speaker B
But I totally skipped the whole classical MCP crap.
Speaker B
So you.
Speaker B
Because you don't then you can actually, if you need to.
Speaker B
You can use MCPs on the fly.
Speaker B
You don't have to restart, unlike unlike Codex or Claude Code.
Speaker B
Where you actually have to restart the whole thing.
Speaker B
I think it's way more elegant.
Speaker B
And also scales way better.
Speaker B
Now you see Anthropic, they do.
Speaker B
They built like a tool call search feature, like something super custom for MCPs.
Speaker B
That was like in beta because it's like so gnarly.
Speaker B
No.
Speaker B
Just have CLIs.
Speaker B
Bot really is good at Unix.
Speaker B
You can have as many as you want.
Speaker B
And it just works.
Speaker B
So I'm very happy that I don't think I got very little complaints about the MCP stuff.
Speaker A
It's kind of back to you're just you're giving it the same tools that humans liked to use.
Speaker B
Yeah.
Speaker A
And not invented stuff for for for bots per se.
Speaker B
Yeah.
Speaker B
Humans no sane human tries to call MCP manually.
Speaker A
Yeah.
Speaker A
You just want to use CLIs.
Speaker B
Yeah.
Speaker B
That's the future.
Speaker A
I'm here for it.
Speaker A
Thank you so much for making the time to sitting down chatting.
Speaker A
It's been a huge inspiration to sort of like when we're texting, you know, over the course of the past couple of years.
Speaker A
And I saw you getting back into the game.
Speaker A
And I was like, Peter, like, what you're telling me?
Speaker A
Like, chase that dragon.
Speaker B
Yeah.
Speaker A
And you were doing like the weird like vibe tunnel thing, etc.
Speaker A
Nobody was paying attention.
Speaker A
And so I'm just like beyond, you know, stoked to see, you know, what's happening.
Speaker A
And, um, and of course, they had to be sort of like a loner from some like tiny country like far away from Silicon Valley.
Speaker A
Just like, you know, bring all of this upon us.
Speaker A
Um, so huge inspiration.
Speaker B
I'm here for it.
Speaker A
Awesome.
Speaker A
Thanks, Peter.
Topics:OpenClawpersonal AI agentlocal AIbot-to-bot communicationautomationAI communityswarm intelligenceAI developmentPeter SteinbergerY Combinator
