Skip to content

AGI ALPHA: A Scalable Substrate for Intelligence Organizations

Latest manuscript · 198 pages · Vincent Boucher · April 30, 2026

Read the complete PDF · Read the canonical Markdown · Inspect the source manifest

This is the manuscript supplied by the repository owner as the current research specification for AGI-Alpha-Agent-v0. Its bytes match the publication in agialpha-first-real-loop. The original whitepaper_v0.1.0-alphav15.pdf remains available as historical context.

The manuscript describes a research program; inclusion here does not establish that every proposed capability has been implemented or independently validated. The implementation map connects the requirements to code, working demonstrations, tests and remaining evidence obligations.

Experience the executable work

  • The Compounding Lab: learn a policy, freeze it, measure transfer to future tasks, account for costs, review the result and export the complete Evidence Docket.
  • Proof Bloom: claims, Proof Debt, native jobs, returned evidence, review and reversible memory.
  • Insight Atlas: opportunity hypotheses, architecture comparisons and evidence trials.
  • Ascension Lab: the earlier paper's connected governance and mission demonstrations.

Publication identity

PDF SHA-256: 4b290d5a8232364b8808c0af8a96e7c9b152afd7ac3f5ba05668a538e073791a.

The supplied PDF, upstream Markdown and all 31 referenced figures are retained without rewriting. The upstream MIT license is included. The versioned manifest pins their source, byte lengths and hashes. The release pipeline verifies them before packaging and checks the public PDF after deployment.