Sughu

@sughanthans1

codex for work. grokbot for life. latest ai news, new model tests and workflows from actual use. still waiting for agi. pinned: best posts + a workflow ↓

following
followers

Odysseys2 launches with 100 web-agent tasks based on real user requests

Carnegie Mellon and Browser Use researchers say GPT-6 Astra led 15 agents, scoring perfectly on 47 of 100 tasks.

8 Sources, 11h ago, updated 11h ago

Featured in

Odysseys2 launches with 100 web-agent tasks based on real user requests

Featured in

Odysseys2 launches with 100 web-agent tasks based on real user requests