Building Pi, and what makes self-modifying software so fascinating

By Pragmatic Engineer

Categories: Product, Startup

Summary

A solo Austrian developer built Pi, a self-modifying AI coding agent that quietly became the engine behind OpenClaw, challenging the industry's obsession with AI-written code by proving quality matters more than velocity—and we need to slow down.

Key Takeaways

  1. Self-modifying software is fundamentally easier to build with AI agents because users can ask the tool to build its own features (like MCP support) rather than waiting for engineers, democratizing feature development.
  2. Non-engineers can now participate in the engineering process without wasting engineer time, but this requires guardrails and process discipline—most companies skip this and ship garbage quality code that users feel in their bones.
  3. AI coding agents and CLI tools are becoming more popular than complex infrastructure like MCP, suggesting simpler, more focused tools win over feature-rich platforms in the AI tooling space.
  4. Building a minimalist, single-purpose tool (Pi) from frustration with existing solutions can outperform feature-heavy competitors—quality and stability beat complexity, even with non-deterministic AI components.
  5. Companies claiming all their code is AI-written are shipping low-quality products; the industry needs to prioritize sustainable development velocity over maximum automation speed.

Topics

Transcript Excerpt

Can we start with the backstory of why you decided to build Pi? I personally like simple tools that are stable that I can rely on even if they have non-deterministic parts. So you can ask Pi to modify itself. Pi doesn't have MCP. People just ask Pi to build MCP support into PI. >> Non-engineers participating in engineering process is a thing now. >> You might have a PM who wants to try out a feature without wasting time of an engineer. Now you can do that. The problem is that people are now so f...