Anthropic shipped the Claude 3 family March 4th, Haiku/Sonnet/Opus, and the flagship posted benchmark wins over GPT-4 on the standard suites: the first time the same-day-launch rival has held the measured frontier, however temporarily, and the market-structure note matters more than the leaderboard (the dance is now a three-body problem, OpenAI/Microsoft, Google, Anthropic/Amazon-Google-money, with Meta’s open-weights flank as the fourth gravitational mass; frontier capability is officially plural, which changes procurement, eval discipline, and the governance math simultaneously: you cannot license a frontier that keeps electing new members). Our own quarterly bake-off (the tooling benchmarks) confirmed the delta on our tasks, which is the only leaderboard the file trusts (the golden-set doctrine): model choice is now a quarterly decision with a regression suite, exactly the commodity-with-switching-costs dynamics this archive filed for clouds a decade ago (the chokepoint homework, now with per-token pricing). The “sparks”-era question (which capabilities arrive at which scale) remains unanswered by anyone including the labs; the eval profession remains the only tractable response; the file remains on message because the message keeps grading correct.
The TikTok endgame opened: the House passed divest-or-ban 352-65 (March 13th, the forced-sale mechanic returning at four-times the 2020 velocity, now with broad bipartisan margin), and the file’s structural note holds from earlier: whatever one’s read on the national-security merits (the classified briefings visibly moved votes; the public evidence remains thinner than the margin), the legislative innovation is platform-specific forced divestiture as a repeatable instrument, the splinternet arriving via congressional markup. Senate timing unclear; the file pre-registers that it passes within the quarter and spends years in courts (the fourth-branch doctrine: everything lands in litigation eventually now).
The sports ledger belongs to the women’s game again: the WSL’s attendance records keep falling (60,000-plus at the Emirates this month; ticket-price telemetry reading like derby data), and by every ratings measurement women’s football is moving past legacy expectations in cultural velocity. The parity thread meets its audience chapter (the graph-import doctrine, organic edition: the product itself as the API), and the Champions League knockouts arrive with the women’s bracket as a main event in the tournament’s history. The file, whose prediction methodology this year seeds by club research-computing spend, could not be happier to witness it.
TIL: constitutional AI’s RLHF variant, Anthropic’s training approach (models critiquing outputs against explicit principles rather than pure human preference ranking): the alignment-technique-as-product doctrine’s second implementation, and the transparency delta (published principles vs. implicit reward models) is itself a governance position (constants-file politics, versioned in public).