The AI productivity split: verifiable vs unverifiable work
danshipper · x · 2026-10-18
leeknowlton's framework: most work splits into AI-verifiable and not, and how you work depends on the bucket. "Verifiable" means AI can independently tell whether its completed work is good. If verifiable, build a verifier or system of verifiers and "let 'er rip" — you're no longer the bottleneck, so tasks scale without your attention. If maybe-verifiable, build an AI judge, but you need evals to test the judge's competence. hammermt adds: the most valuable move now is making an unverifiable domain verifiable by AI, then letting it rip.
Related event: Framework: split work into AI-verifiable and unverifiable tasks(2 posts)→
More from coding & agent
- Anthropic's daily merges jumped from 2,500 to 30,000, so it built its own Git replacement called Scion — Pavan_Belagatti · 2026-10-19
- Open 4B decision model cuts agent costs 81-89% and beats the closed model it learned from — alex_verem · 2026-10-19
- Applied AI practitioner rediscovers industrial engineering: find the bottleneck first — brandon_galang · 2026-10-19
- Open 4B decision model cuts agent costs 81-89% and beats the closed model it learned from — alex_verem · 2026-10-19
- Stardock CEO shows AI-native enterprise workflow: Sharepoint+Jira+Notion rebuilt for AI — draginol · 2026-10-19
- Tencent's Terminal-RSI lifts a 27B model from 51.7% to 86.5% on terminal agent tasks — GabGarrett · 2026-10-19