IA, Not AI
Two philosophies forked from the same machine in 1962 — one automates the human away, one amplifies the human. There is a test that tells them apart, and the industry is failing it on purpose.
Read the piece →Speculative pieces, published unfinished
Every piece here is unfinished on purpose. Each one is drafted half-formed and augmented by AI, not yet reviewed and ratified as finished. So each one wears an honest Unfinished marker. That's not a disclaimer — it's the whole point. Thinking out loud, in public, with everybody, before the ink is dry.
Two philosophies forked from the same machine in 1962 — one automates the human away, one amplifies the human. There is a test that tells them apart, and the industry is failing it on purpose.
Read the piece →Everyone says agent. Everyone says chat. Both frames quietly insist there's a someone on the other end of the line. There isn't — and pretending there is is a category error with consequences.
Read the piece →The big platforms already agree with us: the model is a commodity, the value is your data. That's exactly why they give the model away. We split, completely, on the cure — their servers, or yours.
Read the piece →The model was never the scarce thing — it's becoming a utility, like electricity. The scarce thing is your context, your history, your relationships. The whole fight is over who owns that layer.
Read the piece →Open weights answer the wrong question. The guarantee that matters isn't “we promise not to look.” It's “there is nothing here to look at.” A promise can be broken, bought, or subpoenaed. A fact about where the data lives cannot.
Read the piece →An agent that can act on your behalf has to answer to you — at the moment it acts. The always-listening assistant that acts on what it overhears isn't a helper. It's a parasite. The cure is a design decision.
Read the piece →Doctorow named the decay we all feel: platforms are good to users, then abuse them, then die. The answer isn't a kinder platform. A platform that can trap you will decay. The way out is to not be trapped.
Read the piece →The claim that your data must sit in the middle, on someone else's server, is a business model dressed up as a law of nature.
Read the piece →Some tools make you smarter and leave the gain behind when they're gone. Others make you dependent and leave nothing at all. Which kind we build for the mind is the whole game.
Read the piece →A lost tradition of software built to make thinking better — not to do the thinking for you. It runs from Engelbart through Ted Nelson to Rheingold, and it's worth recovering now, on purpose.
Read the piece →Engelbart to Xerox PARC to the desktop you still use every day — and why reclaiming that paradigm is the work of the sovereign era.
Read the piece →“I,” “here,” “now” do not survive the trip across a network. Every claim should carry who it came from.
Read the piece →A book of faces where the graph and the relationships belong to you, not to a platform — and what that quietly makes possible.
Read the piece →The social theory of intelligence as an engineering program, not a slogan — a civilization of minds, not one lonely genius.
Read the piece →The finished artifact is a flattering lie about how thinking actually works. Publishing before the ink is dry isn't carelessness — it's a method, and here the method is the message.
Read the piece →Everybody's trying to make a Newton. The saner ambition is a Europe. The real quarrel in artificial intelligence isn't about model size — it's a quarrel about what intelligence even is.
Read the piece →Every tool we lean on either leaves us sharper when it's gone or quietly hollows us out. Most AI is being built to hollow us out — and that choice is an engineering decision, not a slogan.
Read the piece →Nobody knows who designed Chartres, and that is not a loss — it is the point. A case for the anonymous cathedral over the author-cult.
Read the piece →The Tetrads — single ideas run through McLuhan’s four laws, printed one to a card.