Topic

Ownership

Who owns the machine, the model, the weights, the cluster.

Software ate the world; capital ate the software. The seven companies with frontier compute do not merely operate AI — they own the means of inference, the training corpora, and the physical infrastructure underneath. Open weights without open compute is a release, not a transfer. This topic tracks the ownership stack from chip foundry to data centre to model registry to application layer, and asks the only question that matters in any economic system: who collects the rent.

Columns (3)

  1. Open Weights, Closed Gates

    Meta's Llama is open the way a company town is open. A column on commons-washing, strategic loss leaders, and the difference between a public good and a loss-making marketing expense.

  2. Primitive Accumulation for the LLM Age

    The scraping of the public web was not a bug. It was the opening move of a classic enclosure. A column on how the training corpus became private property.

  3. Who Owns the Weights

    Model weights are the factory floor of the 21st century, built from scraped public labour and held as private property. A column on the enclosure of intelligence.

Parallels (1)

  1. Then · 1933

    The New Deal forces capital to share with labour the productivity gains of electrification and the assembly line.

    Now · TBD

    Nothing. Gains from AI flow to seven companies and the funds that own them. There is no deal, new or old.

    Productivity without redistribution is not progress. It is theft at a higher clock speed.

Contradictions (1)

Claim
Open-source models democratise AI.
Reality
The weights are free. The 30,000 H100s needed to fine-tune them are not. Democracy with an entry fee is an auction.

Demands (2)

  1. 01

    Public ownership of frontier compute. The means of inference are too important to leave in seven buildings.

  2. 02

    A data dividend. If a model was trained on your words, your face, or your voice, you own a share of the output.

Other topics