Entelligence

I do not have clean answers. What I have is a structure that has held for a while. Beliefs are priors updated in proportion to evidence, and certainty is structurally unavailable, because you cannot step outside your own perception to check. Where cause and effect cannot be worked out ahead of time, the move is designed, survivable experiments with tight feedback, which is a different discipline from having good judgement. And underneath both, the claim I keep arriving at from every direction: a part cannot be validated from inside itself. Local coherence is not evidence of global correctness, and the vantage from which a model can be found wrong is never inside the model.

That last one is why this site has two layers. Thinking holds the standing pages - what I currently hold, why it is stable, and what would overturn it - revised in place and stamped with the date each claim was last checked. Writing holds the dated pieces, each arguing one thing on one day. A post is a point. The standing pages are the curve, and a site made only of points invites you to mistake the most recent one for the position.

It gets tested somewhere real. I co-founded a company that builds software for electric fleets, which is the kind of problem that punishes an elegant model: chargers, duty cycles, grid connections, and a depot manager who needs to know at six in the morning whether every vehicle leaves. Most of the examples are from there.

Nick Butlin - London