P-00 / Projects

Artifacts for governance judgment.

Two projects, selected for signal rather than volume. Each one ties an implementation choice to a governance concept: privacy enforced by architecture, and claims bounded by what the computation can actually decide. Each one states what it does not prove.

Scroll horizontally to inspect all columns.

Project records, artifact types, and current status
RecordArtifactTypeStatus
P-01OpsNormalLocal-first productLive and maintained
P-02Autonomy UndecidabilityBounded verification demoDocumented demonstration

P-01 / OpsNormal

OpsNormal

A daily readiness instrument with deliberate constraints: five sectors, three states, a thirty-day horizon, and no server to send anything to.

OpsNormal dashboard showing a thirty-day status horizon, current streak window, today's sector count, local-only data posture, and the five-sector daily check-in surface.
01

Problem

Readiness and wellbeing tools routinely collect more data than the task requires, and the collection is invisible to the person being tracked.

02

Approach

Constrain the product instead of the policy. Five sectors, three states, a thirty-day window, and browser-local storage only.

03

Evidence

No accounts, no backend, no telemetry. The data posture is displayed in the interface itself rather than buried in a policy document.

OpsNormal daily check-in showing the five readiness sectors and the local-only data posture.
The constraint is visible.The daily check-in shows the limited collection surface directly. The interface states the local-only data posture where the information is entered.

Establishes

A daily readiness tool can operate with no accounts, no backend, and no telemetry while keeping its data posture visible.

Does not establish

Local-first trades sync, multi-device access, and account recovery for user control. Losing the device loses the data unless the user exports it.

Next record: Autonomy Undecidability →

P-02 / Autonomy Undecidability

Autonomy Undecidability

A bounded Python demo and accompanying notes on why universal autonomy verification is undecidable in general, and what that means for how assurance claims should be written.

A small filled region labeled what a bounded check decides sits inside a much larger region labeled non-trivial semantic properties of programs. Annotations state what the bounded check returns, records, and does not establish.
01

Problem

Assurance language often outruns the computation behind it. Verified safe implies a universal claim that no finite check can support.

02

Approach

Use Rice's Theorem to separate the undecidable general question from the bounded questions a finite check can actually answer, then make the bound explicit.

03

Evidence

The demo decides an exact property within a stated step bound and a fixed environment, and records the bound, step count, and assumptions alongside the result.

INSPECTION BOUNDARYThe record exposes the method, bounded claim, and assumptions. The complete implementation remains private and can be reviewed during a hiring process.

Establishes

A finite check can decide an exact property within a stated step bound and fixed environment, with the bound and assumptions recorded.

Does not establish

This is a teaching demo, not a verification tool. It certifies nothing about a real system and establishes nothing beyond the bound.

Review the verification posture / Discuss the work →