Two AI coding agents collided in the same repo — 102 tests passed but changes were incompatible
Time_Positive_9123 · reddit · 2026-11-03
Nick Hill, an engineer at ArcheForge, describes a multi-agent collision: one agent implementing a feature detected files changing underneath it while another process edited overlapping backend routes, tests and config in the same repo, and the first agent stopped rather than overwrite changes. Notably, the backend test suite still showed 102 passing tests — passing tests didn't establish that the two change sets were compatible. With uncommitted work, overlapping edits and unclear ownership, he chose to identify the competing writer and preserve existing work before proceeding. His takeaway: separate AI conversations don't automatically mean separate development environments. He asks how other developers running Claude, Codex or multiple agents in parallel prevent such collisions — Git worktrees, dedicated branches, file ownership rules, or something else.
More from coding & agent
- Factory launches Deployment team, cites Nvidia and Blackstone as users — maddiehfaulkner · 2026-11-04
- Every ran a three-stage AI conference with a 4-person team using Claude, Codex and agents — danshipper · 2026-11-04
- Formally verifying RISC-V CPU mul/div instructions with one AI theorem-prover prompt — satnam6502 · 2026-11-04
- Beyond chatbots: using scheduled Antigravity automations to track upstream SDK changes — rseroter · 2026-11-04
- Rerun open-sources its AI coworkers that run on your ChatGPT or Claude subscription — damnGruz · 2026-11-04
- Gemini CLI PR rewrites PowerShell substitution scanner to fix v0.63 false-block regressions — duvanka-prasad · 2026-11-04