'Keep screenshotting until it hits 9/10': the self-iterating AI workflow people are testing
lxfater · x · 2026-11-15
@chunxiangai argues 'the prompt era is over': instead of micromanaging, let the model run its own loop — 'keep screenshotting until every element scores 9/10, don't exit the work loop.' @lxfater tested the idea on carbon, showing an agent autonomously reading project files and iterating on mobile/desktop UI after a single instruction to 'optimize to 9/10, screenshot each pass, think about the next step' (ran 2m50s).
More from coding & agent
- Perplexity API adds max_staleness to force fresh crawls at $0.001 per fetch — VeryWellVersed · 2026-11-15
- Indie dev shows off T3 Code dashboard with 452 settled agent threads — yihui_indie · 2026-11-15
- Amazon planning World Product Graph to index the web's products for AI shopping agents — jomic01 · 2026-11-15
- Google Labs open-sources 15 AI Design Skills that turn prompts into React UI, 9k GitHub stars — tokumin · 2026-11-15
- 20 GitHub repos turn Fable 5.2 into a full motion studio — 3D, FX, sound, export — PrajwalTomar_ · 2026-11-15
- 79k-Star AI Engineering Course Ships Its Own Agent Tutor for Claude Code and Cursor — ghumare64 · 2026-11-15