Reasoning artifact guide
Where teams should start
Most teams do not need a giant corpus to begin. They need one narrow, explicit, consequential artifact family that can be accepted, rejected, or compared inside a gate.
Docs hub
Customer-facing documentation for the released validator line
This is the curated public docs surface for AI Verification Engine / Tobi Validator: first run, install, GitHub workflow fit, diagnostics, support, and the first reasoning-artifact onboarding materials.
Start with the narrow released CLI surface. Then move into workflow integration, reasoning-artifact discovery, or agent authoring guidance without widening the reading into a broader platform story.
Recommended path
Most teams should follow this order
New here? Start with the released bundle, confirm the local CLI path, then move into workflow integration or reasoning-artifact onboarding only when the first run is stable.
Step 1
Quickstart
Run --help, canon, and golden on the shipped sample
and fixtures.
Step 2
Install and usage
Confirm bundle layout, path assumptions, command usage, and operator expectations.
Step 3
GitHub Actions starter
Move from local success into a narrow CI gate around the released validator CLI.
Step 4
Get 7-day evaluation access
Request a short-lived evaluation token and store it as TOBI_EVAL_TOKEN before your first public workflow run.
Step 4
Finding your first reasoning artifacts
Learn what in your workflow could actually count as a reasoning artifact worth validating.
Step 5
Tsubasa for agents — Stage 1
Use a narrow mental model so agents stop inventing fake syntax and fake runtime behavior.
Reference
Tsubasa Stage 1 syntax and authoring reference
Read the most complete current public-safe reference for Stage 1 Tsubasa syntax, canonicalization, and authoring discipline.
Need help?
Support and issue reporting
Classify the issue and prepare a reproducible report with exact commands and output.
Category explainers
Understand where Organetic fits
These pages explain the category around Organetic without widening the Stage 1 product claim. Use them for first-time readers, AI-search clarity, and founder-led category discussion.
Reasoning artifacts
Reasoning artifact verification
Why traces and evals are not the same as validator-backed artifact verification.
Control Plane
Control Plane for AI agents
Where deterministic validation belongs before action in agent workflows.
Protocols
Agent protocols and verification
Protocols connect agents. Organetic validates reasoning artifacts.
Glossary
Organetic glossary
Public-safe definitions for Tobi, Tsubasa, Liu, Oyu, _h, Stage 1, and Stage 2.
Agent authoring guide
How agents should stay narrow in Stage 1
Agents should not treat Tsubasa as a general scripting language. The winning move in Stage 1 is less noise: explicit accepted cases, rejected siblings, and validator-meaningful case sets.
What these docs now add
They explain what counts as a reasoning artifact, why malformed siblings matter, and why canonical convergence is useful.
What they still do not imply
No runtime/backend maturity, no shipped verification API, and no broad Organetic platform claim.
What to do after reading
Start with quickstart, then move into GitHub workflow fit and your first small repository-owned artifact family.
Next step
Move from docs review to workflow fit and artifact families
Use the docs hub to get started and troubleshoot. Then replace shipped examples with your own narrow artifact family only after the released path is already stable.