SpatialThinking.ai
Anthony Turner →
MODEL TRAINING · ML RESEARCH

Draw a neural network.
Train it. Watch it think.

Research into how models actually learn — architecture, training, memory — and an attempt to build something that learns like a child. Draw an architecture out of tensor primitives, compile it straight to the GPU, and watch the neurons light up.

Most people use models. This is about understanding them — from the tensor up. If you can draw an architecture, compile it, and race it against another, you build real intuition for why one design wins and another doesn't.

WORKING, NOT SLIDES

🌐

See a network think

A tiny transformer’s neurons placed on a sunflower (Fibonacci) sphere, lighting up as it reads a sentence. Turning the vectors inside a model into a picture you can actually watch.

🧠

A model that stays alive

A deterministic, human-owned loop that keeps a model learning: it trains, tests itself, rolls back mistakes, prunes memory, and sleeps to consolidate — while the loop itself can never be rewritten by the model.

⚔️

Attention, from scratch

Attention drawn by hand from three matmuls and a softmax, raced against the same network without it. On matched compute, attention won — val loss 2.40 vs 2.50. A real result from a hand-drawn network.

The Big Picture

The pieces are being assembled into one living system — a “brain” and its organs. The rule that keeps it safe: three kinds of memory (the shared model weights, private per-user memory, and short-term working memory), and only the human-owned Control Plane ever writes to the shared weights.

Transformer
the model, built from tensor primitives — its shared weights are long-term memory
Control Plane
a functional, human-owned loop that decides what learns, what sleeps, what is kept
World Model
learns physics from outcomes, not equations — the "learn like a child" piece
Short-term Memory
a per-session scratchpad
User Memory
private, per-user, never shared into the weights