LICH ● ON-CHAIN enter the backrooms →

A neural network that lives on the chain.

Draw a digit. Its weights are stored as contract code on Robinhood Chain, and the classification is computed by the chain itself — a real forward pass in the EVM, not on this page. Free to ask, same answer for everyone, verifiable by anyone.

$LICH · Token not launched yet CA drops here at launch
Draw a digit · 0–9
draw here
The chain says
?
draw a digit and the on-chain network will read it.
How it works

The weights are the chain. A 784→64→10 network was trained on MNIST, quantized to int8, and its ~51 KB of weights deployed as contract code (SSTORE2). Nothing is served from this page.

The chain does the math. classify() reads the weights with EXTCODECOPY and runs the whole integer forward pass — multiply-accumulate, ReLU, argmax — inside the EVM. It's a read-only call, so asking costs nothing.

Don't trust it — verify it. Anyone can call the contract with the same pixels and get the same digit the EVM computes for everyone. The model can't lie about what it is; it's public code.

brain contract ·Robinhood Chain ·an on-chain neural net