THE PPI BENCHMARK

How well is the world's code actually constructed?

PPI is the Polysemy Preservation Index, a measure of how well code is constructed. It is one reproducible number for how faithfully a codebase honors the full set of requirements it has itself encoded: in its code, in its docs, in the standards it claims to meet, and in the agreements between them.

What the score measures

PPI is not a popularity score or a volume score. It is a resistance score. It asks how much of its own implied, documented, and specified intent a system keeps intact when you read it back, deterministically, in several directions at once.

The same input always produces the same number. There is no model underneath and no randomness, so anyone can recompute the score and get the same result.

The six dimensions

The score is high only when four things line up at once: the code, its docs, whether the two agree, and whether both meet the published standards.

#	Dimension	What it measures	What it catches
1	Code-internal	the code honors its own implied requirements consistently	uneven application (Heartbleed, found at the place the code breaks its own 88% convention)
2	Doc-internal	the docs state their requirements coherently and completely	incoherent, incomplete, or self-contradictory docs
3	Code and doc agreement	code and docs agree on the requirements	drift between the two (docs ahead of code reads as overclaim; code ahead of docs reads as skilled but under-documented)
4	Code vs global standard	the code conforms to the specs and standards it implements	spec non-conformance (Heartbleed, found again via RFC 6520 section 4's MUST discard)
5	Doc vs global standard	the docs meet the documentation standard	sub-standard or non-conforming docs
6	The concordance standard	the global standard for how code and docs should agree	the meta-metric, the bar everyone is measured against

Structure: two internal (self-consistency), one mutual (concordance), two external (conformance), one meta (the bar concordance itself is held to).

The standards are public

Dimensions 4, 5, and 6 reference global standards. doloop publishes those standards, so the bar you are scored against is open and checkable.

The standard for code is the published rule corpus and the PPI logic itself.
The standard for documentation is the doc score: coherence, completeness, and freedom from self-contradiction.
The standard for concordance is the alignment metric: the required fit between code and docs, with an explicit note about which side is ahead when they diverge.

The PPI is the aggregate across all six dimensions. doloop sets the three external bars and then runs every requirement through the checks, so a codebase cannot raise its score with fluent writing alone.

Current benchmark (v0)

Every number recomputes byte-for-byte from the public formula and corpus. The open standard →

loading the benchmark…

The v0 data emphasizes code-internal gaps and early doc, spec, and agreement signals while the full six dimensions are built out across the corpus. The pattern is already visible.