ChatGPT is Breaking Codebases: Devs Complain About Model Regression
wowa93 · reddit · 2026-08-04
A developer recently expressed frustration over a severe decline in the coding performance of ChatGPT (specifically mentioning Sonnet 5.6) while fixing a production bug.
According to the user, the model exhibited contradictory behavior by requesting full files and then immediately asking for snippets. It also failed to follow negative constraints, confidently refactoring functional modules and breaking core features. The developer noted that while it felt like a smart mid-level dev a few months ago, recent updates have ruined its context memory and instruction-following abilities, causing more time spent rolling back Git commits than actually coding.
More from coding & agent
- Survey of 396 Fintechs: 64% of 'AI Agents' Are Just Copilots — LolaCronje · 2026-08-04
- BitBox: Non-engineers successfully audit codebase using AI agents — RSync25 · 2026-08-04
- The Vibe-Coding Curse: AI Impresses Initially but Degrades with Every Detail Tweak — arjunrajlab · 2026-08-04
- Is Traditional RAG Dead? Developers Discuss Six Alternative Architectures — SwordfishWest6860 · 2026-08-04
- Alibaba Amap Open-Sources LongHorizon-Harness: Boosting Agent Task Success Rates via Auditing — aigclink · 2026-08-04
- Indie Dev Builds Universal Memory Layer Across Models: No Vendor Lock-in, Predictable Cost — Opening-Dream9276 · 2026-08-04