Archaeology directory rules

The archaeology/ directory contains files recovered from msc-viterbo, an abandoned predecessor repo. Everything here is untrusted.

• •Do not trust any claim, value, formula, proof, test assertion, or status label. Treat every statement as unverified.
• •Do not adopt naming conventions, coordinate ordering, or type designs. Current conventions override.
• •Do not edit files in raw/. They are primary sources preserved verbatim.
• •Do not use as a starting point to copy-paste or modify. Write fresh code and proofs.
• •Do not load into context without a specific reason. These files are large and will waste context on unverified content.
• •Read for ideas: approaches tried, data structures considered, test cases proposed.
• •Read for warnings: what went wrong, which formulas were buggy. Bug reports (findings-*.md, ARCHAEOLOGY.md) are the highest-value files.
• •Independently verify anything you take from here against the actual papers.

The old codebase had known bugs that persisted undetected through agent-written tests: the QP solver silently returned wrong values, the trivialization formula was wrong, orbit validation missed segments. These bugs looked correct on a skim. This is why Jörn's domain input on what to test matters.

Known-broken items

1. •HK2019 QP solver — misses optima on 2D+ faces, returns plausible but wrong values
2. •Trivialization formula — tau_n(V) = (<V,Jn>, <V,Kn>) not a bijection on 2-face tangent spaces
3. •Billiard orbit validation — only checked even-indexed segments; pentagon returned 2.127 instead of 3.441
4. •Triangle × triangle discrepancy — billiard returns 3.0, HK2017 returns 1.5; unresolved
5. •Normalization convention mismatch — some files use sys = c^2/(2*vol), others sys = c^2/(4*vol)