Plan ordering decides success in 9-week codegen experiment
Ok_Astronomer_526 · reddit · 2026-08-31
A user ran a repetitive code generation task for 9 weeks using the same repo, prompt, and Claude model to generate an OpenAPI client. The only variable was saving each run's execution plan. Comparisons reveal near-identical plans except for one step: whether shared test fixtures are updated before or after call sites. In 6 runs where fixtures came first, the type check passed immediately. In the 3 where they came last, it failed and required a retry. Despite minimal spec changes and no prompt mentions of fixtures, this single random variation dictated the outcome.
More from coding & agent
- Open-Source GitHub AI Code Reviewer: Self-Hosted Models, No SaaS Fees — Arindam_1729 · 2026-08-31
- Library recreates Claude Code, Codex, and Grok interfaces as shadcn components — RichardsonDx · 2026-08-31
- The Post AI Data Stack: Infrastructure Challenges — sh_reya · 2026-08-31
- OpenClaw 2.0 Released: Rebuilds Installation and Browser Experience — heyneighbor · 2026-08-31
- MCP Server Lets LLM Agents Play Civilization VI Natively — jia_seed · 2026-08-31
- Hermes Agent running GLM-5.2 via OpenRouter on a 2GB VPS — granawkins · 2026-08-31