Insights
How we think about building AI.
Occasional notes on rigour, evidence, and what it takes to make AI trustworthy enough for real work.
July 2026·4 min read
The whole system, not a wrapper
Most “AI products” are a general model with a prompt on top. We build the other way around — and here's why it matters.
PhilosophyEngineering
July 2026·5 min read
What “validated” actually means to us
In markets especially, it's easy to produce an impressive backtest and a meaningless one. Here's the discipline that separates them.
MarketsRigor