Anatomy

Diagrams of boxes and arrows describe an idea. These build it. Each piece is the real thing, running: a transformer you can type into, laid out in 3D from its own numbers; a web stack you can overload. The 3D dissections walk through a landmark paper or system component by component, trace every figure to its source, and end with experiments of their own.

10 models 4 AI & ML5 Architecture1 Mechanistic Interpretability 882 parts and components 252 links into the libraries

All10 AI & ML4 Architecture5 Mechanistic Interpretability1

Any Language models 2 Neural network foundations 2 Performance & capacity 1 Web & API systems 1 Distributed systems 3 Hallucination & faithfulness 1

Any Working model 5 Simulation 5

Runs the real algorithm end to end on your own input.

4 models for “numbers”.

Working model · 3D Mechanistic Interpretability · Hallucination & faithfulness

Why 722 becomes 719

A real 0.5B language model opened up at the moment it copies a number from its context: the heads that carry the digit, the votes that choose it, the sampler that sometimes doesn't, and a guard that checks every digit as it is written.

Two people paste the same research notes into a small model and get 722 and 719. A 3D dissection of Qwen2.5-0.5B at the moment it copies the number, with pre-registered experiments on five models: wh…

24 components · Intermediate 18 Oct 2026

Working model · 3D AI & ML · Neural network foundations

Anatomy of the ResNet

Deep Residual Learning, dissected: the paper's own 20-layer CIFAR network in 3D, every feature map and kernel a real one from a model trained on handwritten digits and running in your browser, every number from the paper traced to its sentence.

He et al.'s 2016 ResNet, built in 3D from its own numbers: the paper's ResNet-20 at half width classifies a handwritten digit while the camera walks 27 steps from convolution to shortcut to softmax. …

75 components · Intermediate 17 Oct 2026

Working model · 3D AI & ML · Neural network foundations

Anatomy of Backpropagation

Rumelhart, Hinton and Williams (1986), dissected: the family-trees network of their Fig. 3 at full size, its forward pass and every derivative of its backward pass drawn as real blocks, checked against PyTorch, with a research desk that re-runs the letter's single-run claims across seeds.

The 1986 Nature letter that made backpropagation famous, built in 3D at the paper's own size: the family-trees network answers your query while the camera walks the forward pass, every gradient of th…

50 components · Intermediate 15 Oct 2026

Working model · 3D AI & ML · Language models

Anatomy of the Transformer

Attention Is All You Need, dissected: the whole encoder-decoder in 3D, every matrix a real one from a trained model running in your browser, every number from the paper traced to its sentence.

The 2017 Transformer, built in 3D from its own numbers: a trained miniature of the paper's exact encoder-decoder translates your English number into German while the camera walks through 29 steps, fr…

172 components · Intermediate 15 Oct 2026