Flowers
An open-source personal-agent harness, built to try out reliability ideas. It plans before it acts and asks before anything irreversible.
Lab
Testbeds, side systems, and papers outside our main line of work. Two of the projects run in your browser.
Projects
Systems & demosAedos
Verified symbolic memory for LLMs. It checks the claims in a draft answer against Wikidata and abstains where it cannot find grounding.
Experiments
Autonomous researchVoynich Autoexploration
An autonomous multi-agent loop that probes the structure of the Voynich Manuscript. It ran over 600 corpus-statistical experiments and over 5400 cipher-mechanism simulations.
Adjacent work
Outside the main linePapers by lab members, outside our main line of work.
Preprint
May 2026Cross-Repository Guidance for Coding Agents: A Single-Trial Replication Addendum
A companion to probe-and-refine: can comparable guidance be assembled without ever tuning on the target repo? Mining transferable edits from 50 external repositories yields guidance that resolves 45.4% of SWE-bench Verified vs. 39.6% baseline.
Read →Preprint
Apr 2026