All I wanted was a shitty coding agent that is truly mine. And I’d have loved to just tell you why and how I built pi. But then Peter decided to make it the agentic core of OpenClaw. And now pi is collateral. So yes, this is a talk about pi. But it is also a talk about how agents are destroying OSS, how I deal with that, and a plea to slow the fuck down.

https://x.com/badlogicgames
https://github.com/badlogic
https://www.linkedin.com/in/mariozechner/

AI Engineer

Mario Zechner, creator of the Pi coding agent, shares why he stopped using Claude Code and built his own minimal, extensible AI coding agent harness. He critiques Claude Code and OpenCode for opaque context manipulation, lack of observability, and poor extensibility. Pi is designed around a minimal core with a TypeScript extension API, self-modification capabilities, and full developer control over context. The talk then shifts to a broader critique of over-relying on AI agents: agents compound errors without bottlenecks, produce enterprise-grade complexity quickly, fill spec gaps with low-quality learned patterns, and can make codebases unreadable and unmaintainable. Zechner advocates for scoped, deliberate agent use, reading critical code yourself, and maintaining human understanding of your systems.

Building pi in a World of Slop — Mario Zechner