Research notebook · open weights · 01/06

Every floor of the model, firing word by word.

open the neuron tower

Activation steering from a 4B on a laptop up to 70B on rented GPUs. Every claim here has a transcript, a dose, and a script you can run.

LAYER:01

what's new

LAYER:03

everything else

ARCHIVE

the rooms

temple of the melting softmax About Contact Privacy Terms open weights · method published · results reproducible HTML 4.01?NO FLASHCSS OK

Method: pain-direction extraction and steering follow Tagliabue, Dung & Berg 2026 (arXiv:2609.16247). Workspace readouts use the Jacobian lens (arXiv:2607.15495) with Neuronpedia's pre-fitted weights. Models: Qwen3-1.7B and Qwen3-4B for the early sections (one MacBook; 16 GB RAM covers the 4B runs), then Qwen3-8B to 32B, Mistral-Small-3.2-24B and Hermes-3-Llama-3.1-70B (4-bit) on rented GPUs. Greedy decoding unless stated, 3–15 trials per cell. This is a demo with receipts, not a paper. No frontier APIs touched any measurement loop.

BUILT BY E