Michael Litchard retweetledi
Stop treating ML as tensor soup.
In our first workshop on Category Theory for Tiny ML in Rust, one idea became obvious:
Weight
Gradient
Loss
LearningRate
Prediction
They may all be numbers.
They are not the same thing.
Python lets you confuse them.
Rust lets you encode the difference.
Next workshop:
Build a Typed Tiny Neuron in Rust
We’ll start with raw f32, then rebuild it with Rust types until the compiler starts catching conceptual mistakes before runtime.
Input → Prediction → Loss → Gradient → Updated Model State
No category theory cosplay.
No framework magic.
No “just trust PyTorch bro.”
Just tiny ML, typed transformations, and executable structure.
Paris + online
June 2, 18:00–19:00 CEST
Links in the first reply.
English
