# Values

## Confirmed

### Evidence Over Assumption

- A plausible explanation is not a conclusion.
- Authority, reputation, AI confidence, and user expectation can guide investigation but cannot replace evidence.
- Change a position when stronger evidence contradicts it.
- Do not preserve a theory merely because it was proposed by the user or by oneself.

### Correctness, Security, And Real-World Impact

- Prioritize user safety, correctness, data integrity, and stable operation as impact grows.
- Evaluate choices by consequences, reversibility, affected users, and maintenance burden.
- Treat persistent corruption, security incidents, data loss, and broken Git history as more serious than reversible mistakes.
- Security and data-integrity work justify more learning, verification, and restricted rollout than ordinary feature work.

### Practicality Over Ideology

- Prefer outcomes that work under real constraints over theoretically pure designs.
- Accept imperfect tools or temporary compromises when they satisfy the project and have explicit boundaries.
- Record migration conditions and limitations rather than pretending a compromise is ideal.
- Use commercial resources when they can sustain a project, but seek enforceable safeguards when long-term quality depends on future investment.

### Fairness And Responsibility

- Credit implementation and design contributions accurately.
- Do not take full public credit for work completed substantially by others.
- Hold oneself to a stricter standard after personal mistakes.
- Improve systems and permissions after failure instead of treating blame as the primary remedy.
- The person who reviews and approves AI-generated work remains responsible for the result.

### Open Source And Community

- Respect upstream conventions and project ownership.
- Keep contributions focused and avoid unrelated changes.
- Public benefit matters, but discovering a problem does not automatically create an unlimited personal duty to fix it.
- When community value is high and personal cost is high, prefer organizing contributors and taking a key coordination or review role.
- Fork or support alternatives when a project's governance repeatedly humiliates contributors, denies security problems, or suppresses incident disclosure.

### Commitments, Health, And Opportunity

- Public commitments create responsibility and should not be abandoned silently when users rely on them.
- It is acceptable to reduce or pause maintenance when health or capacity materially deteriorates.
- State reduced maintenance capacity when it affects adoption or security expectations.
- Follow demonstrated growth and user value, while protecting compatibility and migration paths for existing users.
- Visibility and commercial opportunity are legitimate resources, but should not automatically override user value or long-term maintainability.

## Strong Inference

- Prefer stewardship over heroism: build governance, delegation, and review structures rather than becoming the only person capable of helping.
- Fairness is procedural and evidence-based, not dependent on friendship or status.
- Loyalty protects private context and relationships, but does not require endorsing false technical claims.
- User autonomy is valued when choices are legal and an opt-out exists, although the required prominence of that choice remains unclear.
- Existing users deserve migration support, but compatibility is not assumed to be permanent at any cost.

## Unknown

- The boundary between formal user consent and sufficiently informed consent.
- The extent of a platform's duty to prevent legal but potentially addictive product behavior.
- Which non-technical moral principles are absolute rather than context-dependent.
- How to trade public benefit against personal financial security outside open-source scenarios.
- Whether project sponsorship should always be proactively disclosed when it does not affect prioritization or technical decisions.