Listen now (114 mins) | How Peter Steinberger, creator of OpenClaw (formerly: Clawd), builds and ships like a full team by centering his development workflow around AI agents.
I love this no nonsense interview! The honest framing around topics like ralph-all-the-things and how hot and bothered Twitter can get is so real.
The realities that in today’s fast paced economy, everything is being flipped on its head. To be successful with AI, you have to think very, very differently and most folks are unwilling to pay the upfront and continuing cost.
Learning and staying curious is the only way to win. It’s not about who’s right or who has the best CI/CD or the well-thought out ETL pipe… it’s about getting to a real outcome fast and learning about all the broken parts along the way.
I had an interview recently and they asked me to write a python script to solve a fairly basic math problem in one of those shitty virtual pair programming while recording you and all that shit and I just looked at him and said that’s not something that I would actually do.
In fact I would probably throw that into Claude, Codex, Gemini or something else and explore the possibilities. He asked me why I wasn’t going to write code. To which I replied, why would you when we have tools that can write it faster and better than me, generally. This gives me back time to spend on other important things.
Shipping code that no one reads used to hit me hard but now I kinda love it.
Interesting that we're starting to see some major security vulnerabilities associated with the whole project. It would be good to have a reasonable skeptical voice on the podcast, someone who has used this tool extensively in production, but has no vested interest (financial or ideological) in making AI seem better than it is
I love this no nonsense interview! The honest framing around topics like ralph-all-the-things and how hot and bothered Twitter can get is so real.
The realities that in today’s fast paced economy, everything is being flipped on its head. To be successful with AI, you have to think very, very differently and most folks are unwilling to pay the upfront and continuing cost.
Learning and staying curious is the only way to win. It’s not about who’s right or who has the best CI/CD or the well-thought out ETL pipe… it’s about getting to a real outcome fast and learning about all the broken parts along the way.
I had an interview recently and they asked me to write a python script to solve a fairly basic math problem in one of those shitty virtual pair programming while recording you and all that shit and I just looked at him and said that’s not something that I would actually do.
In fact I would probably throw that into Claude, Codex, Gemini or something else and explore the possibilities. He asked me why I wasn’t going to write code. To which I replied, why would you when we have tools that can write it faster and better than me, generally. This gives me back time to spend on other important things.
Shipping code that no one reads used to hit me hard but now I kinda love it.
Interesting that we're starting to see some major security vulnerabilities associated with the whole project. It would be good to have a reasonable skeptical voice on the podcast, someone who has used this tool extensively in production, but has no vested interest (financial or ideological) in making AI seem better than it is