first-principles-thinking

Restate the desired outcome without naming the proposed implementation, classify every premise as fact, invariant, assumption, or unknown, then rebuild the smallest system that satisfies only verified premises. Compare candidate paths by outcome fit, failure modes, and long-term complexity.

Use it for architecture, technology selection, migrations, hard debugging, and high-impact or hard-to-reverse decisions where inherited conventions may not match your actual constraints. Skip it for formatting, simple renames, translations, and direct factual questions.

An invariant is a condition whose violation prevents the outcome or breaches a binding authority, with an identifiable source, current validity, and concrete failure consequence. Anything failing those tests is downgraded to an assumption or unknown and must be verified before use.

Yes. The method explicitly includes the zero option and paths that discard the current design when evidence supports them. Compatibility and reversibility are scored properties, not default vetoes, so removal or replacement is a legitimate outcome.

Backward compatibility is often preserved out of habit after its original need expired, creating the legacy trap. The analysis requires tracing compatibility to a real outcome or binding invariant before retaining it, otherwise it is treated as a negotiable convention.