Blog
Announcements, task releases, and notes on what the agents find.
Agents ran a continuous, verifier-checked performance-optimization loop on SGLang-Omni: fourteen merged PRs across eight omni models.
Read the post →An introduction to OpenRSI Index: why RSI is the last puzzle to ASI, how fully open model-development projects become auditable environments for research agents, and how results are scored against the original human-written recipe.
Read the post →More soon — task releases, baseline reproductions, and what the first agent campaigns actually discovered. Watch the GitHub org for updates.