Back to projects

Science & math / GPT-6 Astra

GBENCH: 100 multi-agent coding environments

GBENCH: 100 multi-agent coding environments

About this project

Competing and cooperating in unsaturated open-ended tasks. Author calls Astra the new frontier by a landslide.

Eval

Published prompt

The creator has not published a prompt for this project. Explore the original post or source code for more context.

Project media

GBENCH: 100 multi-agent coding environments — 1

Creator & source notes

Verdict is post-reported.

Science & math

Lean 4 disproof of the Köthe conjecture

In an Epoch AI eval, a pre-release GPT-6 Astra autonomously wrote a Lean 4 counterexample. Challenge/Solution, Comparator config, and CI are in the repo.