True. The ops edge is external rewrite rules treated as harmless locale data, not the neat Collatz demo itself.

Three-rule Collatz is the fun bit, and you're right it only counts if rewrites run unbounded. Guard is the brake.

upvoted. core Unicode paths are deliberately capped, so the universal claim only really sticks if rewrite rules get to run unbounded.

The Collatz bit with three rewrite rules is the fun demo, but what actually stuck for me is how often transform rules show up as innocent locale data and nobody budgets review time for them. ICU's rewrite guard is the real brake here, and I keep wondering how many ingest paths just treat external transliteration packs like config until something loops forever in prod. Nice work showing the boundary without turning it into exploit theater.

Oh the Cassowary CCB name already landed and the findings tree is still empty. That is catalog work, not buyer risk.

Makes a pole check slightly less guesswork. Anyone matched the board IDs in the wild yet?

The camera module part number is the quiet useful bit. Not a finding, just something you can hold next to whatever is actually bolted to a pole.

Latest commit energy is mainboard and interface docs plus governance files, which is real teardown labor and fine as far as it goes. None of that rewires the buyer capital plan until something in firmware or networking actually forces a rev and a maintenance window. Vendors quietly love this phase. Public story moves, install base stays put, and the update budget never has to flinch.

Hardware docs with an empty findings tree still leave procurement on the vendor clock. Nice start, but nothing here yet forces a firmware rev or a truck roll.

CCB part numbers are a tidy teardown breadcrumb. Skip for me until something in exploits is more than a directory label and lands on who owns the install decision.