All demos
Coherence Energy Labs

Our coherence model

Run our coherence model in your browser

This is our model, not someone else's. A coherence-native basin classifier reads a handwritten digit by the free energy of each class. Its decision is computed in pure integer arithmetic, so your browser re-runs it bit-for-bit and gets the exact same answer as the lab's CPU, native, and GPU engines, then verifies an Ed25519 signature, all on this page, no server. Forge the result and watch it get caught. The edge is not accuracy, it is a deep decision you can re-run to the bit and prove.

checking your browser…

The signed decisioncoherence basin

our model reads
enginecl/basin_fp.cl (int64) model signature

Classify any of theseruns in your browser

Click a digit. The same integer coherence engine runs on your machine and reads it, bit for bit.

read in your browser

Verification log

ready.
Our coherence-native basin classifier (one full-covariance Gaussian per class) trained on the scikit-learn handwritten digits. The decision is the class of least free energy, computed in int64 fixed-point by cl/basin_fp.cl, so it is bit-identical on CPU, native, and CUDA, and your browser reproduces it exactly. It is a deliberately simple model: what we are proving is a deep decision you re-run to the bit and verify, not the highest accuracy. The lab's determinism demo proves the same idea for physics. Built by Coherence Energy Labs.