WHAT NEW IN X11.02 MON

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

gpt-6POST

Clipping Bench tests whether AI agents can earn money, Astra finishes with $9,514

A new benchmark measures earnings rather than puzzle solving: each agent team starts with $300 and three social accounts, clips long livestreams and posts for 30 days with pay per verified view, and GPT-6 Astra ends with $9,514 while Sol ends with $5,707.

rob.
@robiartecArchived 23 days agoOriginally posted 2026.10.10
rob. @robiartecOriginally posted 2026.10.10
Open original on X

Agent benchmarks usually measure whether they solve a puzzle. Few measure whether they make money. Clipping Bench does the opposite: → each agent team starts with $300 and three social accounts → picks campaigns, cuts clips from long streams and posts them over ~30 days → brands pay per verified view, and the score is the final cash GPT-6 Astra ends with $9,514 and GPT-6.1 Sol with $5,707

This post quotes another post: x.com/usenaive/status/2108592968596935013

x.com/robiartec/status/2108605916710855091Archived 2026.10.10 10:03