Skip to content

Repository files navigation

Capability Lab banner

Capability Lab

Evidence-grounded capability modeling under explicit governance boundaries.

CI Documentation Python 3.11+ Source available: PolyForm Noncommercial

Observation is not evidence. Evidence is not a claim. Supported is not mastery. Current is not permission.

Documentation · Architecture · Constitution · Consumer boundary · Reference map

Status: stable research subsystem · public source-available checkpoint · development is demand-driven.

Capability Lab is an experimental Python framework for representing what has been observed, what the evidence can support, what remains uncertain, and what development paths may be worth considering — without turning a model, score, recommendation, or product view into authority over the person.

The current public distribution uses the PolyForm Noncommercial License 1.0.0; commercial use that requires rights beyond that grant requires separate commercial licensing.

Why Capability Lab

Most learning and professional systems lean on proxies: courses, titles, credentials, tests, or self-reported skills. Capability Lab explores a different architecture built around provenance, explicit uncertainty, append-only evaluation history, human review, and governed state transitions.

The system deliberately preserves distinctions that are easy to collapse:

observation
!= evidence
!= interpretation
!= claim
!= evaluation
!= derived state
!= persisted state
!= accepted state
!= current state
!= progression advice
!= product/read projection
!= permission or authority

It does not attempt to compute a person's value, intelligence, destiny, professional license, or right to act.

Current architecture

The generic public path composes end to end:

flowchart LR
    O["External observation"] --> H["Human review"]
    H --> E["Neutral evidence"]
    E --> I["Governed interpretation"]
    I --> C["Claim"]
    C --> V["Conservative + domain-sufficient evaluation"]
    V --> D["Complete-history capability state"]
    D --> P["Persistence"]
    P --> A["Explicit acceptance"]
    A --> S["Current-state selection"]
    S --> G["Advisory progression + current profile"]
    G --> R["Governed product/read snapshot"]
Loading

The architecture treats the read surface as a projection, not as authority:

PRODUCT / READ SNAPSHOT
!= CURRENT-STATE SELECTION AUTHORITY
!= PROGRESSION AUTHORITY
!= CAPABILITY UPDATE AUTHORITY
!= PERMISSION OR PROFESSIONAL AUTHORITY

Executable contracts cover the generic observation path through governed current state and the product/read boundary. The product surface has no capability write-back authority, and automatic closed-loop capability updates remain not authorized.

Stable consumer boundary

The current consumer contract is CurrentStateGovernedPlayerWindow.

A consumer may present governed current state and advisory progression. It does not choose the current state, grant readiness or mastery, or turn a serialized snapshot into live authority.

Start with Consume the governed read boundary.

Research lineage

Capability Lab began with Civilization Bootstrap Engineering as a demanding domain stress test. The repository retains that research and Pilot history, while the generic path is deliberately independent of the Pilot-specific production path.

The current public Git history intentionally begins from a clean source-available snapshot. Technical PR/milestone references retained in research documents are documentation provenance, not published Git ancestry.

See Publication lineage and the reference/archive map.

Development

Requires Python 3.11+.

python -m pip install -e ".[dev]"
python -m pytest -q

Documentation development:

python -m pip install -e ".[docs]"
zensical build --clean --strict

Reporting and discussion

Use the repository Issue Forms for reproducible non-sensitive bugs, documentation problems, and research/architecture discussion.

Security- or privacy-sensitive reports must follow SECURITY.md. Do not put credentials, person-scoped records, private captures, private workspaces, or exploit details into a public issue.

Research discussion and conceptual proposals are welcome. Until a dedicated contributor-rights process exists, substantive third-party authored material is not accepted for inclusion and cannot be merged; see CONTRIBUTING.md.

License

Capability Lab is source-available, not OSI open source, under the PolyForm Noncommercial License 1.0.0 (PolyForm-Noncommercial-1.0.0).

Commercial rights are available separately where a license from the project is required; see Commercial licensing.

Earlier versions were previously distributed under Apache-2.0. Rights already granted for those earlier copies remain in force for those copies; see License history.

About

Experimental framework for evidence-grounded human capability modeling, governed evaluation, progression, and safe external observation.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Contributors

Languages