WHAT NEW IN X11.02 MON

AI collections from X · 22 collections · 65,263 posts

jevVIDEO

Sonnet 5.5, Dots and Jev ship an agent in 55 minutes for $4.92

A six-layer, bottom-up agent build uses Sonnet to write the policy and tools, Dots for data and 240 test cases, and Jev to cap and lock it down, passing 92% at a token cost of $4.92.

Woody
@woody_researchArchived 27 days agoOriginally posted 2026.10.06
Woody @woody_researchOriginally posted 2026.10.06
Open original on X

SONNET 5.5 + DOTS + JEV SHIPPED AN AGENT IN 55 MINUTES. THE AGENCY QUOTE WAS $120,000... the receipt says $4.92 WRITE → sonnet 5.5 drafts the policy, the prompts and six tool integrations FEED → dots brings the data, the retrieval and 240 test cases TEST → all 240 run once: 221 pass, 19 fail FENCE → jev caps it and locks it down SHIP → deploy goes last, never first six layers, bottom up. no layer starts before the one under it exists the part that matters: the 240 cases and the 90% bar were written before the build. it cleared 92%, and the 19 failures sit in a file, not in a footnote the quote: 800 hours × $150 policy and prompts alone: 120 hours, $18,000 the real run: 1.14M tokens, $4.92 nobody paid the $120k. it's just what someone would charge you building got cheap. keeping it true didn't: $59,160 a year, $37,440 of it a human reading 270 tickets a week the build is the cheap part now ↓

This post quotes another post: x.com/0xkiyoro/status/2107135100911563017

x.com/woody_research/status/2107141156567916672Archived 2026.10.06 10:04