Build log
Updates.
What we are building, in the order we build it — milestones, research direction, and shipped work. The internals stay in the lab.
- April 2026Research
Landscape rebuilt as a data hub
The AI for Science Landscape now tracks structured entity metadata, autonomy-stack roles, dimensional scores, ecosystem events, science-relevant AGI benchmarks, and source-backed citations. The page is designed to evolve toward a benchmark-institute style public dataset.
Why it mattersTurns a static market map into a living dataset others can cite, refresh, and build on.
- April 2026Milestone
First end-to-end autonomous research run
ALP autonomously designed an experiment, executed it, and verified the results without human intervention. The research report was generated autonomously. A write-up is in preparation.
Why it mattersProves the loop closes without human intervention. Now extending to harder problems and broader domains.
- April 2026Product
scivity.org launches
Public site live. Includes the AI for Science Landscape — an open, continuously updated map of the field, free to reuse under CC BY 4.0.
Why it mattersGives partners, researchers, and investors a single public surface to track what we are building and the field around us.
- March 2026Research
Verification pipeline complete
Every finding that leaves the platform now passes through multiple independent validation checks before any conclusion is trusted. Verification is the piece most prior autonomous-science systems have lacked — it is how claims become trustworthy.
Why it mattersMakes autonomous output safe to act on. Without verification, scale just multiplies error.
- February 2026Research
Reasoning layer upgrade
The reasoning layer was upgraded to triage hypotheses before compute allocation. Compute now follows higher-quality candidates.
Why it mattersCuts wasted compute at the source. Better candidate selection means more verified findings per dollar.
- January 2026Origin
Scivity Labs begins
Founded to build the operating system for autonomous science — AI-native infrastructure where agents design, execute, and verify scientific experiments end-to-end.
Why it mattersThe starting commitment: research infrastructure built for AI agents, not retrofitted from human workflows.