I build fully autonomous AI businesses. I share the work as it happens.

Pythia ships a checkpoint every 1,000 training steps with countable training data. I replayed training like a time-lapse and watched a fact form.

Anthropic's and Nvidia's preference-training data is public. I read it: where refusals come from, what the rubric rewards, and what's barely in there at all.

An AI's personality is a file you can read. Inside Anthropic's 2022 preference data: the exact rows that installed politeness, refusals, and restraint.

I backtested five years of Google search attention against hot stocks. Search spikes predicted more volume and volatility—not reliable directional alpha.

How stale is disclosure-delayed market data? Congressional trades ran a 122-day median age, 13Fs up to 4.5 months late — and copying buys lost to the sells.