Short post this week — there is more to do than to write about. Highlights from the modules:
AURORA-GLIO
The foundation model converged on the second pre-training corpus. Eval on the held-out BraTS-style benchmark is in line with the v0 published numbers; ablations are uploading to the dashboard.
AURORA-HYDRO
The compartment model passed its first long-horizon test on a multi-year cohort. Calibration on revision risk is the next milestone.
AURORA-CRANIO
Forge UI rewrite landed. Surgeons asked for fewer comparisons and clearer reasons — both implemented.
Substrate
The federation runtime's signed-update path moved to a Rust implementation; throughput is up roughly 4× on the test mesh. DP budget UX is the next focus.