# Mechanical Ethics

Mechanical Ethics is a human-facing framework for examining consequential
decisions under uncertainty, especially where formal correction can arrive
after a threatened path has hardened.

## Read this

Choose either form of the same current released reader:

- [`MECHANICAL_ETHICS.md`](MECHANICAL_ETHICS.md) - inspectable Markdown source;
- [`MECHANICAL_ETHICS.pdf`](MECHANICAL_ETHICS.pdf) - typeset 6 x 9 inch reader edition.

Both present the current **released Mechanical Ethics v0.8.0 baseline**. The Markdown is the editable source; the PDF is a deterministic generated carrier, not a separate manuscript.

| Released form | Bytes | SHA-256 |
|---|---:|---|
| Markdown v0.8.0 | 100,253 | `e3267c3e41e7fdfb44c9a9279250a1a7a38d5b9c389a594fef40cfea75fdbe76` |
| PDF v0.8.0 | 450,125 | `e9ca37b249b4d915b18e954b6c72ebd999ae4be8ceed2c0aaf395922d0c6b56f` |

The corrected Figure 2 source/carrier are pinned at:
- SVG: 3,001 bytes; SHA-256 `54bb7822db06493d9af86a56c225d5b0713881fa7e634300985d665fb9cfde17`;
- PNG: 44,180 bytes; SHA-256 `9f6681bf1ccaf5a4516ac55ebf1278a38064cae0d044ee164794857d45c4ae71`.

```text
RELEASED / FORMAL BASELINE / CURRENT READER / NOT VALIDATED
```

The four diagrams in [`figures/`](figures/) are explanatory carriers, not evidence.

## Current status

Mechanical Ethics v0.8.0 is the current released formal baseline. It promotes the reviewed v0.8.0-beta4 reader after multiple hostile/cold AI review rounds, bounded source repairs and deterministic carrier checks.

The release adds no new claim of validation. The project deliberately stopped general prose revision after mirror falsification found diminishing returns from further review-loop refinement and stronger pressure for real-world testing.

Mechanical Ethics v0.7.0 remains preserved as the previous released baseline. Earlier versions, evidence records and build history remain recoverable through Git history and the dated `branch-archive-20260829-pre-minimal-surface` tag; they are not part of the current reading surface.

External comparison with [FPF](https://github.com/ailev/FPF) and later stronger-owner review narrowed contribution claims and reinforced the project's integration posture. They did not earn conformance, endorsement, a compulsory combined workflow or proof of practical advantage.

## What remains unresolved

- standing under protected-scope uncertainty;
- hard collisions and competing clocks;
- grounding and ordering of protection floors;
- representation and legitimate decision authority;
- closure where the beneficiary is also judge;
- residue triage under finite repair capacity;
- enforcement, brake and interruption authority;
- usability for ordinary people without unusual reserves.

Release decision: **23 September 2026**. Promotion changes release/baseline status and document control only; it does not establish validation, universal applicability, authority, efficacy or resolution of the open ethical questions.

## Review, history and licence

Critical review is welcome through the repository's GitHub issues. Agreement
is elaboration, not validation.

The existing [AI training permission](AI_TRAINING_PERMISSION.md) remains scoped to the owner-controlled released v0.7.0 material described there. It does not automatically extend to v0.8.0. Release of v0.8.0 is not, by itself, a new training or reuse licence.

No general reuse licence is granted beyond that specific permission. Public
visibility alone is not reuse permission. Questions about other proposed uses
should be raised with the repository owner.

```text
RELEASED_BASELINE != VALIDATION
EXTERNAL_AGREEMENT != VALIDATION
COMPATIBILITY != CONFORMANCE
DIAGRAM != EVIDENCE
PUBLIC_VISIBILITY != REUSE_PERMISSION
```