LLM writing still garbage in 2026, dev shows the blunt prompt he uses
arthurcolle · x · 2026-10-22
A developer complains that LLMs are still terrible at writing in 2026. The attached screenshot shows his workaround: a forceful prompt telling the model to drop LLM-style language, pull up to 60 skills from GitHub for serious writing, and generate a PDF within 6.5 minutes.
More from coding & agent
- a16z hosts SpaceX leaders on what it takes to deploy AI agents in enterprises — stuffyokodraws · 2026-10-22
- MCCL Paper Reframes Agent Memory as an Active Driver of Continual Learning — Dongfang Li · 2026-10-22
- GRSE Assigns Behavioral Credit in Multi-Agent Teams via Group-Relative Comparisons — Mayoclinic · 2026-10-22
- Harness Scaling Trains Agentic RL Across Diverse Configs for Better Generalization — RUC · 2026-10-22
- Claude Code autonomously searched user's laptop to find a screenshot they forgot to attach — menhguin · 2026-10-22
- LaunchMotion ships 208 launch-video Skill scenes so AI agents storyboard and render product trailers — yihui_indie · 2026-10-22