NUROSIS

NUROSIS NOTES

Notes from the build

Numbered working notes, citation-backed, written as we go. Research first; conclusions only where the numbers hold.

  1. 01

    Agents learned to act. They still can't listen.

    Two years of progress gave AI agents the ability to execute. Nobody gave them a faster way in. That missing input layer is a brain interface, and it is the reason Nurosis exists.

    4 AUG 2026AI AGENTS · INPUT LAYER · BCI

  2. 02

    The dataset nobody recorded

    Every brain-to-text decoder learns from examples of signal matched to words. Almost nobody has those examples at scale, because hospitals never asked language questions. A brain-to-text company is, first, a data-collection company.

    18 AUG 2026DATA · EEG · RESEARCH

  3. 03

    Ten bits a second

    The measured size of the human output bottleneck. Everything we build starts from one number and one ratio.

    1 SEPT 2026BANDWIDTH · NEUROSCIENCE · INTERFACES

  4. 04

    The signal under the hair

    What an EEG electrode actually measures, why the skull is the villain of this story, and what that physics means for anyone trying to read words off a head.

    8 SEPT 2026EEG · PRIMER · SIGNAL

  5. 05

    How a model learns to read a brain

    Windows, representations, tokens, and the foundation-model bet, explained without the math but with none of the meaning lost.

    15 SEPT 2026MODELS · PRIMER · DECODING

  6. 06

    What an agent does with a decoded thought

    Decoding gets words off a head. Agents turn those words into actions without becoming dangerous, and the whole trick is knowing when to ask.

    22 SEPT 2026AI AGENTS · PRIMER · SAFETY

  7. 07

    The map on your head

    Why EEG electrode names look like map coordinates, why every measurement needs a reference point, and which neighborhoods the speech network lives in.

    29 SEPT 2026EEG · FIELD GUIDE · MONTAGE

  8. 08

    The brain's frequency atlas

    Delta through high-gamma: what each rhythm actually reflects, how the signal gets from scalp to samples, and why speech work lives at the fast end.

    6 OCT 2026EEG · FIELD GUIDE · SIGNAL

  9. 09

    How to read a decoding paper

    Character error rate versus word accuracy, within-session traps, and the small print that separates a real result from a demonstration. A field guide to the claims.

    13 OCT 2026BENCHMARKS · FIELD GUIDE · HONESTY

  10. 10

    Wet and dry: the electrode question

    The physics of the skin-electrode interface, what gel actually buys, what dry electrodes trade away, and why the choice is a data strategy, not a comfort feature.

    15 OCT 2026HARDWARE · FIELD GUIDE · EEG

SUBSCRIBE VIA RSS