The Model Didn’t Get Dumber. My Agent Skills Got Stale.
When Claude Opus 5 and GPT-5.6 arrived, I expected my coding agents to become noticeably better. Instead, some of my workflows felt worse. The agents seemed more eager, less predictable, and occasionally “dumber” than before. Naturally, I blamed the new models. Very scientific of me. Maybe it was a skill issue Then I watched Andrej Karpathy’s interview on the No Priors podcast. One idea stuck…
The article discusses the author's experience with Claude Opus 5 and GPT-5.6 models, where they expected significant improvements in their coding agents but instead observed negative impacts. The author attributes this to outdated skills and prompts that were no longer compatible with the newer models. After conducting an audit, the author discovered that the instructions were written around the behavior of older models, leading to redundancy, unnecessary steps, and overuse of certain tasks.
The author then cleaned up the prompts and retested the skills, resulting in noticeable improvements in performance. The article concludes with a checklist for maintaining custom skills when transitioning to new models, emphasizing the importance of regularly auditing and updating these instructions to ensure optimal performance.
Brief written by urgent.news from Dev.to's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
