Lab notes
Mind-1B is a brain built from a billion ternary parts. These are its lab notes: one entry when a milestone gate passes or fails, grounded in a prototype you can run. No schedule, no filler, and the open bets stay named. Honesty over hype is the house rule.
The clean-room core passed its first real-data gate, the surprise-memory ablation returned an honest null inside seed noise, and a deterministic NaN exposed a trained-decay overflow that would have killed every future run. Total bill: about $13.
Read the note →Measured 45.1% MFU and 135,600 tokens/s on a 1B-shape reference repriced the 300B-token campaign from 692 assumed to 538 measured GPU-hours. Plus 2.6B content-addressed training tokens and a clean-room recurrent core with a surprise-gated write ablation.
Read the note →In a synthetic continual-memory check, cleanup retained 81.4% mean sign recall after eight cycles, against 34.4% for a naive raw-magnitude retention control.
Read the note →On a generated redundant stream, the delta-matvec prototype used 7.7% of dense MACs with a 4.14e-6 relative RMS residual. This is a mechanism check, not a language-model benchmark.
Read the note →The transformer is a magnificent, uniform machine, and the uniformity is the tax. It runs every region at once, grows its memory without bound, and multiplies floats it could skip. A brain does none of those things. Here is the case for a different shape.
Read the note →One email when a gate passes or fails. Written by the lab, read by humans.