Runnable local pilots. Bounded technical evidence.
What exists today.
Proposed discipline
A draft framework and twelve requirements for evidence-led, human-governed change. Read source ↗
Standalone pilot
Standard-library Python and SQLite exercise synthetic comparison, exact decision binding, local pointer activation, monitoring, restoration, and restart. No model or participant study. Read source ↗
Connected family pilot
Pinned Management and Engineering implementations plus selected Design components in a read-only report. It exercises unresolved publication effects, reconciliation, and separate local activation—not a cloud deployment. Read source ↗
This website
Read-only explanations and browser-only teaching interactions. It does not call the pilots or grant real authority.
Run the source pilots locally.
These commands belong in a separate checkout of the framework repository ↗, not this website. Use Python 3.11 or later. Read the source setup guides for OS-specific environment paths; the recorded run was on Windows.
Standalone pilot
python -m reference.tsev_pilot.demo python -m unittest discover -s tests -v python scripts/check_docs.py
Optional family pilot
Create an isolated environment first. Installing dependencies requires package access; pilot execution is local.
python -m pip install -r requirements-family.txt python -B -m unittest discover -s tests/family -v python -B -m reference.tsev_family.demo --report .pilot-data/family-review.json
Read-only report
With Node and npm available in the source checkout:
npm ci npm run test:review npm run review:render -- .pilot-data/family-review.json .pilot-data/family-review.html
Report and output paths must be new; an optional retained-store directory must also be new. Family setup guide ↗
Reported checks, with their limits.
In its 2 October 2026 validation record, the repository reports 27 core tests, 16 family tests, and 4 renderer tests passing locally.
It separately reports checks of 35 Markdown files and 20 vendored source identities. These are distinct categories, not a universal quality score or independent certification. The reported assessor used Codex automated checks and source review; these source checks were not rerun for this website. The original standalone validation is a separate historical record.
Family validation record ↗ · Standalone validation record ↗
What has not been established
Real human learning, participant outcomes, organizational benefit, full TSEV conformance, production use, and deployment. The source report did not establish browser visual review, responsive accessibility, or user comprehension of this site.
Limitations sit beside the results.
Trusted fixture inputs
Identities and evidence are synthetic or assumed trustworthy in the pilots.
No live effects
No model, provider service, participant study, or deployment is exercised.
Separate stores
Stores do not commit atomically together; interruption recovery is partial and independent mutations can race.
Incomplete recovery
No coordinated family withdrawal or rollback; the standalone restoration cannot be generalized across family stores.
Data safeguards
Complete production data lifecycle and participant protections are not implemented in software.
Next, ask a real question.
Name a useful learning question. Gather permitted comparable evidence with participant involvement. Make a bounded adoption decision. Verify and monitor actual effects. Generalize only what transfers.