Join waitlist

Build the next chip.

Describe a medical device and what it must detect. AI scientists design the chip inside it.

Taken apart.

Sensor signals in, a small fine-tuned AI model in the middle, one prediction out.

Smaller, then smaller.

AI scientists fine-tune thousands of candidate models at once and redraw the chip around the best.

Gates in this layoutIllustration
21,904first layout
Each line is one candidate model learning; lower is better. Weak ones are stopped early. The amber one is the smallest that gets under the line.

The model, as gates.

The winning model is compiled into logic gates, with its weights wired in.

Checked, bit for bit.

The circuit runs every test window and must give the model's answer each time.

Model and chip, side by sideIllustration
3,880 of 3,880
test windows where the circuit matched its model, bit for bit
2,330
logic gates in the finished circuit

Early feasibility test, Sat 21 Nov 2026, on a laptop. Not yet clinical.

Every chip, a device.

Each chip the environment designs is the core of a different medical device.

IllustrationDevices chips like these could run.

About

Three people building the .

  • Adrian Mladenic Grobelnik

    The AI-scientist environment and the GPU batch.

  • Antonio Aciob

    The interface: the stages, the live curve and the chip view.

  • Oran Barak

    The clinical side: which errors matter.

What will you build?

Name the device and what it must detect. The AI scientists do the rest.

Join waitlist