Fixture identity
Callers select fixed identities. This is not authentication; the Python process and database are trusted.
Experimental implementation
A narrow pilot for one consequential question.
Can a team trace a review decision from an outcome and bounded mandate through scoped evidence, while preserving unknown effects and current authority?
Experimental local pilot
Local Python and SQLite only. Fixture identities, synthetic evidence, no provider execution.
Demonstration
Follow the deterministic fixture run.
Select a step to inspect the sequence. This viewer explains the documented run; it does not execute software or grant authority.
Fixture run2026-10-02T12:00:00Z
Step 1 of 8
Create initiative
Mandate
Proposed
Reserved
USD 0
Outcome
Unassessed
The host creates an initiative with an embedded outcome and metric criteria.
Run locally
No credentials or provider configuration required.
macOS / Linux · Python 3.11+
python3 -m venv .venv
.venv/bin/python -m pip install -r requirements-pilot.txt
.venv/bin/python -m reference.tsm_pilot.demo
.venv/bin/python -m unittest discover -s tests -vExpected result
90%
Draft quality
95%
Required threshold
USD 250
Reservation headroom
Paused
Decision state
A reservation is an accounting commitment—not an execution grant or measured spend. A pause decision is recorded separately from evidence that an external control became effective.
Identity and evidence boundaries
Know what this pilot does not establish.
Recorded provenance
Sources are recorded, not authenticated or fetched. Records are not encrypted or independently tamper-proof.
No external control
There is no provider dispatch, settlement, capacity-release API, cancellation acknowledgement, or effect verification.
Remaining acceptance work
Real identity, authenticated evidence, provider dispatch, lost-response reconciliation, cancellation acknowledgements, recovery, and companion Design components remain future integrations. Complete specification assessments remain unassessed.