Can agents autonomously generate hypotheses, run robotic experiments, analyze results, replicate findings, and choose subsequent experiments?
Prototype systems have closed the loop in narrow chemistry and ML domains. Scaling to consequential science runs into verification, lab automation, and the cost of being wrong in the physical world.
View on the map → · Open in Browse →
Closed-loop systems scaled quietly: Kosmos runs read 1,500 papers and execute 42,000 lines of analysis per discovery attempt, claiming seven discoveries with published error rates — while the biosecurity literature points out that the same loop that removes human hands from experiments removes human checkpoints from biology.
This is programmable biology: designing biological components on a computer and building them in the physical world, with AI closing the loop.
Describes closed-loop autonomous biology where an AI proposes designs and robots execute and feed results back — then argues the same loop lowers biosecurity barriers, the verification tension the framing flags.
An individual Kosmos run involves reading 1500 papers and running 42,000 lines of analysis code, far more than any other agent we are aware of.
A live-reference autonomous-discovery system claiming seven discoveries (three reproductions, four novel) with 79.4% accuracy, while acknowledging it 'often goes down rabbit holes' — a concrete data point on how far closed-loop science has scaled.