Sep 9, 2026
Six weeks of the Abuja beta: what the first reads taught us
Notes from the clinics running GRX Edge day to day — where it helped, where it didn't, and what we changed.
img
Notes on the technology and the work — how the models read, what the clinics teach us, and the research behind the diseases GRX Edge looks for. Written as it happens, in plain language.
Why the next generation of edge imaging AI is moving away from heavy, cloud-dependent statistical models toward neuro-symbolic AI — and how Rad, at 1,455 parameters, reads a chest film entirely on-device.
Read the pieceNotes from the clinics running GRX Edge day to day — where it helped, where it didn't, and what we changed.
CAD is recommended and deployed — yet it serves the patients who need it most least well. Two compounding gaps: who the models were trained to read, and who can actually reach them.
Read the pieceExtending the same core infrastructure from chest to mammography — what carried over, and what didn't.
The failure modes that never show up in a benchmark, and the ones we had to build around before anything worked.
Why local prevalence changes what a reading should say, and how the engine adapts without being retrained from scratch.
No cadence promises — we write when there's something worth saying. Leave an address and we'll send it.
We only use it for journal updates.