Mollick on Opus 5.5: prompt less, build in checkpoints instead of one-shotting
eldonredwards · x · 2026-10-30
Addy Osmani argues most people are overprompting Claude: newer models like Opus 5.5 need a single clear goal prompt plus instructions on how to verify completion. Ethan Mollick adds that it's often better to build in checkpoints where the model reports interim ideas or prototypes, so you can guide and give feedback rather than one-shotting the goal.
Related event: Stop Over-Prompting New AI Models, Engineers Say(2 posts)→
More from coding & agent
- Codex appears to default to GPT-6 Luna for memory extraction and GPT-6.1 Terra for organizing — TheMoonMidas · 2026-10-30
- Factory launches Droid 1.0, a new engine for its coding agent targeting complex projects — sonofalli · 2026-10-30
- E2E-SWE benchmark: 186 tasks test coding agents building repos from scratch, best pass@1 just 67.7% — jyangballin · 2026-10-30
- LLM-driven pentest framework ARTEX exposed: one FOFA query finds 300+ live panels — cyb3rops · 2026-10-30
- How one builder runs his life on a Grok Bot agent: skills repo, finances, home lab — Rasmic · 2026-10-30
- Qwen Code nightly: Generated UI on by default, WebShell message paging — qwen-code-review-bot · 2026-10-30