ANNA

ANNA 1.3 orchestration

ANNA coordinates independently contextualized agent tasks, controlled admission, competing proposals, dissent and evidence records. The recovered 1.3 source passed all 101 automated tests and strict TypeScript checking in an isolated local copy on October 13, 2026.

These are local software checks and simulated or mocked workflows. The separate proposed live model-backed experiment remains unapproved. Authorized API spending is $0. Credentials and genuine provider behavior remain unverified.

ANNA Neural Intelligence

A standalone PyTorch convolutional network trained from random weights on real image data, with held-out evaluation and measured CPU inference comparisons. It has its own source, tests, checkpoints and evidence.

Integration into ANNA is future work and requires standalone review and a separate authority design. This model does not replace the orchestration system.

Verification boundaries

• Locally verified: orchestration tests, neural training, CPU evaluation and profiling.
• Simulated: ANNA's existing non-provider workflow simulations.
• Planned: reviewed controlled integration and testing on additional hardware.
• Unmeasured: Tesla silicon, GPU execution and production inference.
• Public release: the owner approved this evidence page; the standalone neural source and measured evidence are public on GitHub.