Dev finds asking LLMs for prompt and tool design advice often hurts results
mimi10v3 · x · 2026-11-14
Developer mimi10v3 complains that when she asks agents for advice on designing tools and prompting them, things tend to go worse rather than better compared to doing it herself. She proposes building an LLM that is good at prompting LLMs — one that could estimate task difficulty and token requirements and suggest good designs for tools and data.
More from coding & agent
- Epic AI MegaJam: Build a Game in 4 Days With Unlimited AI Agents for $65k Prizes — chrisfirst · 2026-11-14
- AI-designed main menu concept art for 'The Rat: Director's Cut' — chrisfirst · 2026-11-14
- agent-browser v0.43 cuts install size to 17MB and startup to 2ms — cramforce · 2026-11-14
- Open-Source Tool Jaull Scans Your PC and Recommends Local LLMs That Actually Fit — Ton-Llop · 2026-11-14
- Free 95-minute deep dive: a Google engineer on building reliable AI agents — goyalshaliniuk · 2026-11-14
- LISL whitepaper: local 12B model completes 38+ work units in under 9k tokens — MammothTreat6751 · 2026-11-14