The DeepSeek aftershocks settled into the pattern the file pre-registered: Nvidia recovered a meaningful fraction within the week (Jevons winning the argument, every hyperscaler’s earnings call this fortnight raised capex guidance, citing inference demand; the $600B deletion now reads as a volatility event in an intact thesis, the belief-uptime doctrine surviving its sharpest test), the R1 methodology is being replicated and extended across every lab and university cluster (the replication machine, now for training recipes, distilled reasoning models are appearing weekly, and the open-weights lag has functionally inverted on cost-per-capability), and the policy layer responded on schedule: export-control tightening proposals, congressional DeepSeek hearings, and government-device bans (the TikTok playbook, re-run at model velocity, the geography now treats weights as the border-crossing artifact). The Paris AI Summit (this week, the Bletchley lineage’s third act) completed the vibe shift with unusual honesty: safety-institute language downgraded, “opportunity” language promoted, the US vice president explicitly warning Europe against regulation, and the US and UK declining to sign the communiqué, the governance-velocity file records its first reversal: the summit series born from frontier-risk consensus now photographs an acceleration consensus, eighteen months later, with the same attendees (the governance-by-exit doctrine operating at the state level: the concerned institutions didn’t lose the argument; the argument’s venue moved).

The Champions League’s new knockout-playoff round delivers its marquee tie tonight: Real Madrid against Manchester City, the empire against its counter-architecture, a fixture the group chat has been prediction-journaling all week (the win-probability discipline armed, the methodologies degenerate, the popcorn provisioned). The file pre-registers nothing beyond the structural note: every dynasty meets the architecture built to beat it eventually, and which side of that sentence each of these clubs occupies is precisely what the two legs will grade. Fifteen days will file it.

Work dispatch: our Q1 model-portfolio review (the quarterly bake-off, now a committee with a budget line) executed its first DeepSeek-era rebalance, reasoning-tier workloads split across three providers plus a self-hosted distilled model for the latency-tolerant tier (the cost collapse arriving in our own P&L as a 60% inference-budget reduction on unchanged quality, per the golden sets), and the principal-file logs the strategic note for the archive: model plurality plus open-weights viability has converted “which lab” from a marriage into a portfolio, the chokepoint doctrine’s happiest possible ending, provided the evals stay honest (the discipline that makes the portfolio possible; the file’s decade bet on boring infrastructure pays again).

TIL: MoE (mixture-of-experts) inference economics. Activating parameter subsets per token decouples model size from serving cost, the architectural key under DeepSeek’s pricing and increasingly everyone’s roadmap: the cost-of-thought margin war now has a supply-side answer, and “how big is the model” has quietly become the wrong question entirely (the twelve sentences gain a corollary: the expensive thing gets computed by the cheapest expert that can).