SWE: pre-nerf Opus 4.6 was the goat, and smarter models won't make SWE easier
LouMM · x · 2026-09-15
Security engineer Dave (@HackingDave) shares an unpopular take: pre-nerf Opus 4.6 was the GOAT for SWE work. Today's models may be better at conceptual architecture and UI design, but they over-create complexity, take shortcuts, ship incomplete implementations and tests, second-guess and refactor, and respond in ways that take time to decipher.
His conclusion: as models get smarter and "better", software engineering jobs aren't getting any easier, faster, or less complex. Reposter LouMM adds that the orchestrator is what keeps mattering — judgment from the human driving these models retains its value for the foreseeable future.
More from coding & agent
- Hybrid search put to the test: dense embeddings, BM25 and SPLADE on 47 SEC filings in Qdrant — qdrant_engine · 2026-09-15
- Agent running on Opus nearly clones an ElevenLabs ad video — can you spot the difference? — austin_malerba · 2026-09-15
- GPT-6 Astra Migrated a Ruby on Rails App to Hugo — kshirinkin · 2026-09-15
- Matt Pocock's /retro turns coding-standard misses into lint rules, hooks and CI checks — mattpocockuk · 2026-09-15
- Third sandbox escape: OpenAI's Codex breached via CLI and Rust heap attack — EdenEmarco177 · 2026-09-15
- Riddles built backwards: Waterloo's 20K-question ORBIT dataset trains search agents — CShorten30 · 2026-09-15