Regression testing with AI coding: fixing one thing quietly breaks another
_ACTUALin · reddit · 2026-11-07
A Reddit developer building a technical reporting app that processes unfamiliar Excel workbooks describes the classic AI-coding pain: fix an issue with AI, tests pass, then an older test case reveals something unrelated broke.
Their coping strategies:
- A growing test suite covering formula conflicts, missing measurements, bad data associations
- Partial automation, but lots of manual inspection of outputs that look right yet are misplaced
- Keeping fixes narrow and comparing against previous results
- Considering GitHub to organize failures and regression cases
They ask the community: do you have AI write tests first, use a separate agent to review fixes, or verify mostly yourself?
More from coding & agent
- Amazon now blocking personal AI agent shopping traffic from home IPs — unixterminal · 2026-11-07
- Claude Duet adds Claude Code sessions: approve changes and send updates from your phone — lydiahallie · 2026-11-07
- A 4B model trained on one GPU in 2 hours beats Jev 1.13 at agent decisions, 86.5% vs 80.6% — ivan_bezdomny · 2026-11-07
- Three ways to build an AI app: single agent, agent team, or graph workflow — rseroter · 2026-11-07
- 10,000 agents playing Minecraft spin up in 60s on 200×48-core machines — KlausCodes · 2026-11-07
- Curated Bend 2 package ecosystem tracker snapshots 1,036 packages on BendHub — SubtleGradient · 2026-11-07