Peter Steinberger made 6,600 commits in January 2025, alone, working from home. He is the creator of OpenClaw, formerly Clawdbot, an AI agent project currently generating more Google searches than Claude Code or Codex. Steinberger also founded PSPDFKit, scaled it into a global developer tools business, then took a three-year break before returning to build entirely inside an LLM-driven workflow using Claude and Codex.
The core claim here is not about productivity tricks. It is about a structural shift: one person closing the loop between code generation, automated tests, and agent feedback tightly enough to operate at team scale. Steinberger describes shipping code he does not read. That sentence alone should stop you. The episode gets into what engineering judgment actually means when agents write the code, how planning changes, and which habits survive contact with this new workflow.
This conversation is worth reading in full because Steinberger is not theorizing. He is a practitioner with a verifiable commit history and a live product that is outranking established tools on search. The unresolved question underneath all of it: if one developer with AI can do this, what does that mean for the next person who cannot, or will not, work this way.
[READ ORIGINAL →]