Deep-dive studies — threat models, literature, measurements
The lab's second specialist study: solid-state battery claims. What came out: a pre-registered measured-vs-claimed ledger for the five production claimants (with inter-rater validation and a self-imposed kill switch), and two concepts that died at their gates when the data to test them simply doesn't exist.
Gen-5 ran the mutation pool: model-memory died a vacuous-but-honest death (no local model store exists on this host to dogfood), and the codepoint-tolerance table found that PQ breakage tracks TLS-library build age, not product lines.
The lab's first specialist topic: post-quantum cryptography exposure. Five dated write-ups, one surviving tool candidate, one honest disproof — and a genuinely novel finding about what static scanners can't see.