Senior Dev Finds Claude Opus 5 Excels at One-Shots but Struggles with Legacy Codebases
markusn42 · reddit · 2026-08-14
A senior software engineer shares observations from using Claude Opus 5. They note that the model has historically been frustrating for large, existing codebases, often struggling with technical debt and conflicting documentation.
However, when tested on a brand new project (a non-trivial Chrome extension scraping utility) in high-reasoning mode, Opus 5 performed exceptionally well. It set up the repo architecture autonomously, made 17 commits in a single session, and produced working code that passed an adversarial review by Codex.
The engineer speculates that Opus 5 might be specifically bad at complex projects with legacy debt but excels at clean-slate, one-shot tasks, suggesting that current benchmarks need to cover this aspect better.
More from coding & agent
- Terminal Night 3 talk slides released, missing demo video to follow — 4310sy · 2026-10-16
- Agent developers are sitting on unused API credits that can exceed actual usage — EnvironmentalDay8864 · 2026-10-16
- Notion's bet on agents: abstract away deployment so users and agents just write code — simonguozirui · 2026-10-16
- Four Gemini 4 Argon agents beat Minecraft's Ender Dragon autonomously in 27 hours — Maxence Faldor · 2026-10-16
- Tencent's Terminal-RSI builds recursively self-improving terminal agents, hitting 19.84% on Terminal-Bench 4.0 — tencent · 2026-10-16
- Tsinghua's AutoSFT uses multi-agent collaboration to automate SFT data research — TsinghuaNLP · 2026-10-16