Launch-season December, fully institutionalized (#277’s counter-programming cadence now has a holiday calendar): OpenAI is mid-“12 Days of Shipmas” (o1 full release with a $200/month Pro tier — the #285 inference-economics file notes the price point as the margin war’s opening bid: thought, tiered; Sora’s public release Monday — the #271 demo now a product, with the provenance stack #263 shipping alongside as C2PA metadata, graded “necessary, insufficient, present” by the file), and Google counter-launched GEMINI 2.0 Wednesday (explicitly framed as “for the agentic era” — Project Mariner browsing autonomously, Astra seeing through cameras, and the #267 watch-list item now the STRATEGY SLIDE of both giants simultaneously: 2025 is being pre-announced as the year models stop chatting and start DOING, and the #250 agent-spring’s expensive recursion meets its second, better-funded spring). The file’s eval-discipline note (#260, always): “agentic” converts error rates from per-ANSWER to per-ACTION-SEQUENCE — compounding failure probabilities across multi-step tasks is the reliability math the demos elide (a 95%-per-step agent is a 60%-per-ten-step agent), and the orgs that instrument sequence-level golden sets FIRST will own the trust layer the way #135’s rehearsers own reliability. Pre-registered for 2025’s grading.

Google’s OTHER launch mattered differently: WILLOW, their quantum chip, demonstrated below-threshold error correction — logical qubits whose error rates DROP as physical qubits are added (the #166 file’s “watch error-corrected logical qubits, currently approximately zero” checkpoint, formally passed eight years later) — plus a benchmark septillion-year classical-comparison claim the file discounts per the #166 IBM-rebuttal doctrine (the goalposts will move; they should; that’s the sport). The honest summary for the shelf: fault-tolerant quantum computing graduated this week from “if” trending toward “when,” with “when” still measured in years-to-decade for cryptographically-relevant machines — but the #077 crypto-agility homework (post-quantum migration inventories; NIST’s standards shipped this year) converts from prudent paranoia to scheduled maintenance, and the file notes that the industry that took four years to patch Heartbleed (#031) now needs a decade-scale key-rotation it should have started yesterday (the #275 dependency-tiering RFC gains a cryptographic appendix; the invoice #146 always arrives).

The month’s ledgers otherwise: the #286 California file’s federal echo — the incoming administration’s AI-policy signals (the EO’s #289 fate clarifying toward repeal-and-replace; a “czar” appointment from industry) keep the #253 geography redrawing, and the archive keeps minutes; Devin-class coding agents and their benchmark controversies continue the #250 taxonomy’s maturation (the gap between “solves scoped tickets in demos” and “survives our repo” remains the #260 golden-set’s favorite chasm); and the sports ledger notes Mbappé’s massive contract details resetting the #288 contract-engineering meta within six weeks of its filing (the term-sheet arms race compounds; the file’s twelve-sentences doctrine #289 adds no thirteenth).

TIL: quantum error-correction thresholds — the surface-code crossover where adding physical qubits HELPS instead of hurts (noise below threshold compounds correction faster than error): the entire field was a bet on reaching one inequality, and Willow’s data says the inequality is real. Some walls are load-bearing until one measurement; the #258 replication machine now gets its decade with this one.