David Fowler: build CLIs for agents to explore systems and fix bugs autonomously
mariorod1 · x · 2026-10-12
David Fowler shares his latest engineering practice: building CLIs that let agents explore systems.
Core idea:
- Not deterministic end-to-end tests, but a headless way for agents to exercise common system interactions
- Once available, ask the agent to explore all the ways things can go wrong — it gets real feedback from running the system, not just reading code
The full recipe:
- Access to the code
- A harness to quickly verify changes
- Telemetry to understand what's happening
With all three, an agent can find a bug, fix it, and verify the fix on its own — including via interaction combinations you never thought to test. He says this pattern emerged while building Aspire, the open-source agent-ready tool for composing distributed apps.
More from coding & agent
- "Why learn Rust when you can write C++ and ask Claude for bugs?" dev joke sparks debate — wordgrammer · 2026-10-12
- 10 open-source GitHub repos to cut your AI agent's token bill — victor_explore · 2026-10-12
- Personal software will be huge — and tastemakers will be the new value — alexmacgregor__ · 2026-10-12
- Gauntlet Loop: The Prompting Method Behind Claude's 55k-Line Call of Duty Build — majidmanzarpour · 2026-10-12
- User reports two Codex resets and one Claude usage reset this month — rudrank · 2026-10-12
- "A full day's work two years ago": AI agent ships production-grade upload flow unprompted — tristanbob · 2026-10-12