A local rule. The next byte.
Watch a small, attention-free neural model generate template text one byte at a time.
Causal byte language
Select a compute backend and start the engine.
No server, account or model API required.
Awaiting a prompt.
Ready when you are. Nothing is running yet.
Small context. Visible computation.
A causal neural rule reads only the current byte and left neighbours. Each update stage has its own predecessor cache. Dense operations use the selected compute backend; Python nonlinearities and sampling run inside ZIPP WASM.
Trained on template text, not a general-purpose language model. No fast-memory branch or online training.
Read the research context ↗Reproduce & inspect the source ↗Inspect the Python that runs in ZIPP
Actual bundled source. Matrix graphs use the selected backend; this viewer is read-only.
Download language.pyStart the engine to view the verified source, or use the download link above.Explicit backend. No silent substitution.
WebGL2 needs supported hardware and floating-point render targets. CPU reference is an intentional fallback, not a GPU result.
Live state, not a benchmark claim.
Colours visualise actual state. Per-command timings include host overhead and readback; they are not hardware comparisons.
Original checkpoints. Bounded sessions.
Python runs in a worker with execution limits. Stop engine terminates the worker and discards the current state.