Agora launches Git-based shared memory for AutoResearch agents as 13 workers improve a frozen model without training or gradients

Agora launches a Git-based shared-memory system that lets 13 AutoResearch agents improve a frozen 119.6M-parameter model without gradients or training data, producing 1,703 immutable DAG-recorded contributions and a reproducible 145-commit recipe.

• Agora launches a Git-based shared-memory system for AutoResearch agents, recording results and verifications as immutable append-only DAG commits.
• In a nearly 12-day weight-transfer run, 13 language-model workers publish 1,703 contributions and reach 1.899 bits per byte on a frozen 119.6M-parameter hybrid without training data or gradient updates.
• The winning 145-commit recipe draws 165 independent reproductions with no failures, but the community converges on a bigram-recipe monoculture until a human intervention redirects exploration.