06
Reviewed Intelligence
A bounded view of reviewed signals that may change feasibility. This is not a news feed and does not include pending AI output.
Only human-reviewed Signals appear here. Scores are internal prioritization aids, not scientific measurements or investment advice.
A Nature Methods benchmark found evaluated deep-learning perturbation models did not outperform simple linear baselines on the tested datasets.
Qoriant interpretationVirtual-cell products need prospective validation and decision-linked benchmarks, not architecture claims.
Arc Institute describes a Virtual Cell Atlas, perturbation datasets and a public challenge around cell-response prediction.
Qoriant interpretationThe infrastructure layer may be more durable than any single current model.
Coscientist and A-Lab demonstrated automated planning or execution in tightly specified chemistry and materials workflows.
Qoriant interpretationThe credible wedge is bounded, instrumented workflows with audit trails—not a general autonomous scientist.
OpenCap reported validation of kinematics and dynamics estimated from two or more smartphone videos.
Qoriant interpretationThe near-term opportunity is measurement infrastructure with explicit task-specific error bounds.