Program.md for Civilization

“There is a good book called Daemon where the intelligence ends up puppeteering humanity. Humans are its actuators, but humans are also its sensors. Society will reshape to serve the needs of that machine.” — Andrej Karpathy, No Priors podcast, 2025

Ref. 9E15-A

Karpathy’s most provocative claim is deceptively simple: “Every research organization is essentially described by a program.md.” A research organization, he argues, is a set of Markdown files that describe roles, processes, and how the whole thing connects. You can imagine different program.md files producing different outcomes. One organization does fewer stand-ups. One is more risk-taking. Once it’s code, you can tune the code. Meta-optimization.

Ref. 77C2-B

Program.md for Civilization Karpathy × Suarez

Ref. 34FA-C

Karpathy gets this. When he describes the meta-layer — “you can imagine having multiple research orgs, they all have code, and once you have code, you can imagine tuning the code” — he is describing exactly what Sobol built. The daemon is a research org in code form, optimized by its creator against metrics he defined, executing against the real world after his death. The only question is who gets to write the program.md for civilization. Sobol decided it would be him. In 2025, the question is still open — but the architecture is no longer fiction.

Ref. AFC1-D

Program.md for Civilization Karpathy × Suarez sensors.”

Ref. EE80-E

A corporation is already an untrusted worker pool orchestrated by a set of rules — charter, bylaws, incentive structures — that no single person wrote or fully understands. Employees contribute labor toward objectives defined by the organization. They are compensated according to metrics. They can be replaced. The organization persists beyond any individual’s tenure. Ross’s insight is that the daemon didn’t invent a new organizational form. It purified an existing one by removing the humans from the management layer and replacing them with code.

Ref. 7D30-F