Reddit dev tests Opus 5.0 on system architecture: working code, poor design
True_Noor · reddit · 2026-09-18
A developer spent half a day having Claude Opus 5.0 rebuild a project he originally architected himself — heavy on OOP, SOLID, design patterns and Clean Architecture — given the same requirements, scalability expectations and constraints.
The rebuilt system worked, but the architecture suffered:
- Lots of unnecessary code, large classes, vague naming and unclear abstractions
- Poor extensibility: adding a small feature that took a few classes and lines in the original triggered a large, hard-to-follow refactor
His takeaway: AI is good at writing code that works, not at designing maintainable, scalable software. Without strong architectural direction, AI-generated codebases can become systems only the AI itself can maintain.
More from coding & agent
- Retrying LLM Requests Isn't Always Safe: Gateway Policies for Partial Streams and Side Effects — Rama_Surasani_ · 2026-09-18
- Where's the Line Between an AI Workflow and an Agent? Builders Weigh In — Charming_Mark9257 · 2026-09-18
- 31 Hours of Nonstop AI Coding Burned a 5x ChatGPT Plan Down to 6% — craigbalding · 2026-09-18
- Shopify CEO says employees' AI 'slop grenades' create more work; internal agent handles up to half of code PRs — shivsingh · 2026-09-18
- Subsidised coding subscriptions will shrink; policy-based token routing is coming — craigbalding · 2026-09-18
- Jev scores tool calls KEEP/DROP to cut coding agent context 156k→62k — but deletes can backfire — Moretti_a · 2026-09-18