Practical AI Humility in the Age of Agentic Coding
124 snips
Mar 17, 2026 
Steve Klabnik, a Rust contributor and co-author of The Rust Programming Language, experimented with AI agents to build the Rue programming language. He shares his shift from AI skeptic to hands-on experimenter. Topics include agentic coding, using AI to accelerate language design, tradeoffs in developer workflows, and trust and safety when letting agents modify code.
AI Snips
Chapters
Books
Transcript
Episode notes
Hands-On Agents Changed A Skeptic's Mind
- Steve Klabnik shifted from AI skeptic to practitioner after trying agent tools like Claude and VS Code agents.
- Hands-on use revealed agentic workflows were materially different and actually useful compared to early chat-only experiments.
Mom's Boyfriend Used ChatGPT To Draft Reports
- Steve describes showing ChatGPT to his mom's boyfriend who used it to draft many reports and then fact-check them, gaining productivity by treating generated text as an editable draft.
- The non-programmer's workflow highlighted tolerance for imperfect output plus post-editing as a practical pattern.
Programmer Determinism Vs Everyday User Tolerance
- Programmers expect determinism; everyday users accept non-deterministic outputs and build habits to verify them.
- That mindset gap explains why many developers dismiss LLMs while normal users still find them valuable.
