About the Program

The NEO AI Research Program is the publishing arm of NEO AI. It publishes the argument behind the platform, the registers that constrain what the platform is allowed to claim, and the corrections to both.

It exists because of a structural problem the industry has not solved: an AI platform's marketing surface is written faster than its evidence accumulates, and there is no mechanism that makes the gap visible. The registers here are that mechanism. They are published, versioned and machine readable, and every capability claim on any NEO AI surface resolves against them.

What it publishes

Research papers, technical notes and institutional publications, each with an identifier that is issued once and never reissued. A capability register. A controlled vocabulary. An errata record. And the arithmetic that bounds what the Program is permitted to claim about efficiency.

What it refuses to publish

A benchmark it has not run. A comparison it cannot reproduce. A number without a rung behind it. A capability status that is not in the register. A generated summary of a paper that reports no measurement.

Editorial standard

Every material statement carries a claim class. Every capability statement carries a status from the register. Corrections are numbered, dated and published rather than edited in silently, because an estate that quietly deletes its own defects cannot be audited.