Notes from building the studio and its systems — what we’re making, what we’re learning, and what we’re letting loose in Wildlab.
Field Notes · July 2026
Stop Buying AI Features. Start Redesigning the Work.
Nearly every New Zealand business is now “using AI” — yet productivity went backwards. The problem was never adoption. It’s that switching a tool on isn’t the same as redesigning how the work gets done. Where the returns actually come from, and where to start.
Wildlab · July 2026
Rendering four billion years of continental drift in the browser
An in-house experiment: a React Three Fiber globe that morphs Earth’s continents through deep time — from a molten Hadean surface to today’s mapped coastlines — driven by hand-written GLSL and a palette tuned to each era’s climate.
Wildlab · June 2026
A local-first personal AI that never phones home
ElliotOS pulls from eight-plus personal data sources — health, calendar, email, productivity, news — and writes a morning and evening digest to Slack. The AI runs locally on Ollama, so none of that data has to leave the machine.
Wildlab · April 2026
We replicated a viral “61% from Trump tweets” trading claim. It didn’t survive.
An open-source project claimed a 61.3% hit rate predicting S&P 500 direction from Truth Social posts. We reproduced the number exactly — then tested whether it was actually tradeable on Polymarket. Zero of a hundred pre-committed strategies survived.
Wildlab · April 2026
Can a laptop trader beat the bots on Polymarket’s AI markets?
AI-themed prediction markets are the fastest-growing category on Polymarket — ~$10M of volume across 234 markets in a month. We pre-registered a study into whether a person with a laptop and $1,000 can still find edge the bots miss. Here’s the setup, and one thing we already can’t explain.