What we build, and what it found.

Polarize's design system for scientific interfaces: tokens, zero-build CSS, evidence badges, shadcn/ui + visx React components. Live Storybook ↓

• Post · 2026-09-28 Redesigning the homepage, one version at a time The homepage went through several versions, each one kept. The first tried a familiar product-site layout and the third threw out every rule. The one that shipped took the first one's spacing, cards, drawings and numbers onto a white page...

Evidence-first literature RAG for small local models: code controls search, evidence, verification and the run log

• Post · 2026-09-24 A RAG tool that keeps the model in check The model is used for bounded transformations, while code controls search, evidence, verification, abstention and the run log.

KIT Adaptive Preregistration: shared agent protocols (preregistration, conversation logging) installed into every repo with the kit_ap CLI

• Post · 2026-09-28 One question per study: laying out research across repositories A study is one question and everything used to pursue it. Its findings live in one shared corpus, its predictions are tagged before anything runs, and nothing in it resolves a path on one person's machine. The scaffold writes that...

EEG/ECG/EMG with an Arduino + Olimex SHIELD-EKG/EMG: firmware, bridge, recording workbench, playback, BIDS export

• Post · 2026-08-25 Can a $70 board tell me I'm wrong? A single-channel EEG rig characterised in numbers: in-band 40 Hz power across six captures, why broadband RMS is the wrong figure to specify a rig by, and a proposed test of analog bandwidth as a validation axis.

Falsification benches for how a memory of a sound could be held in neural tissue: spiking-network and habituation models, with preregistered experiments. Mostly negative results.

• Post · 2026-09-25 A spiking network built to lose: what it would and would not remember In simulation, synaptic depletion never kept a sound-specific memory, and only plasticity gated by the receiving cell did. A published receptor model drained untrained responses under a protein-synthesis block, where the real cell does not.