LLM ideation experiment: just ask for non-obvious ideas — and Opus beats smaller models by a wide margin
arjunrajlab · x · 2026-10-16
arjunrajlab shares empirical findings from testing LLM ideation strategies:
- The best prompt is the plainest: simply asking the model to come up with non-obvious ideas beats poems, random words, personas and other free-association tricks for novelty.
- A judge/filter step helps: having AI rank its own ideas surfaces good ones, and the AI's picks match the author's own judgment of what's good.
- The Opus vs Sonnet/Haiku gap is stark: charts show Opus alone can reliably turn strategies like "list the obvious, then avoid it" and "critique and revise" into 10+ distinct useful-novel ideas out of 80, while Haiku trades usefulness for novelty and Sonnet sits in between.
A directly actionable methodology for AI-assisted brainstorming and research ideation.
Related event: Asking for Non-Obvious Ideas Boosts LLM Creativity Most, Study Finds(3 posts)→
More from coding & agent
- ThePrimeagen mocks AI coding: infinite tokens still produce an Electron app — seatedro · 2026-10-17
- Generative UI 101: Building Live Interfaces with Agents on Opus 5.5 and Rivals — EXM7777 · 2026-10-17
- MCP server review: gate tool writes on auth/traversal findings, not test score — SamuelFmedeiros · 2026-10-17
- DeepSeek shows rapid vision gains: agent self-checks screenshots to build 3D scene — cedric_chee · 2026-10-17
- He gave an AI agent phone and email access to collect a $500 unpaid invoice — curious_vii · 2026-10-17
- AI:AM transcript: agent lockouts, hidden valence, and formal proofs for Linux kernel bugs — The Cognitive Revolution · 2026-10-17