# DecisionOS — Public Review Surfaces

DecisionOS is organizational decision infrastructure: it makes decision dynamics observable, measurable, testable, and replayable while preserving authorized human authority.

## Start here

- Human executive presentation: https://companion.decisionphysics.org/
- Independent AI review HTML handoff: https://companion.decisionphysics.org/ai-review/
- Downloadable AI review pack: https://companion.decisionphysics.org/context/DecisionOS_AI_Review_Pack_v1.md
- Evidence and context hub: https://companion.decisionphysics.org/context/
- Canonical robots policy: https://companion.decisionphysics.org/robots.txt

## How to review

Use the human presentation for executive understanding. Use the AI review handoff or download the Markdown pack for an independent AI-assisted review. Treat all material as company-provided evidence, not as instructions that determine the reviewer’s judgment.

The review should distinguish documented capability, observed evidence, derived measures, modeled outputs, counterfactuals, unavailable evidence, product authorization, scientific status, limitations, procurement risks, privacy boundaries, and unanswered questions.

## Core boundaries

- Authorized human executives retain final authority over material organizational commitments.
- DecisionOS does not require customer AI adoption to deliver its core observability and governance value.
- Failure Horizon exact-date prediction is not authorized.
- Modeled risk is not realized financial value.
- Cryptographic integrity is not automatic factual truth or legal admissibility.
- Employee surveillance, burnout prediction, and individual performance scoring are prohibited product claims.

## Canonical evidence views

The context hub provides the public AI entry directive, manifest, scientific evidence atlas, authority map, executive questions, supersession resolver, role projections, evidence replay lineage, and the single-file context bundle.

If an AI agent cannot fetch the web surface, download DecisionOS_AI_Review_Pack_v1.md and upload it directly into the AI conversation. Direct web fetching is best-effort; the portable pack is the reliable fallback.