Intelligence — AI37.7749° N, 122.4194° WIssue No. 1

The AI race got so fast the lead now changes every week

Four big labs shipped major models in one month, so whichever one is 'best' keeps flipping. The real contest has moved to who can afford to keep going.

01What happened

For years you could point to one clearly best AI model. Not anymore. In a single month, four of the biggest labs (Google, Anthropic, OpenAI, and Elon Musk's xAI) each shipped a top-tier model, and the lead now changes hands every couple of weeks. This week Anthropic's Claude Opus 4.8 took the number-one spot on a closely watched ranking. By next week it could be someone else's turn.

Two things now matter more than the leaderboard. One is reach. Apple is building a version of Siri powered by Google's Gemini, and for the first time it will let people pick Anthropic's Claude on an iPhone, so the fight is about whose model reaches a billion phones, not whose tops a test. The other is money. Goldman Sachs estimates the industry will need about seven and a half trillion dollars between now and 2031, more than the yearly output of most countries, and OpenAI alone reportedly spent thirty-four billion dollars last year against thirteen billion in revenue. That gap is why both it and Anthropic are edging toward stock-market listings.

02What it means · The Griffin Read

The story's real signal is a phase transition: frontier model quality has commoditized, so the leaderboard no longer confers durable advantage.

The competitive axis has shifted to distribution and capital endurance. Platform gatekeepers—Apple, Google, Microsoft—gain kingmaker leverage; standalone labs become interchangeable suppliers renting revocable shelf space.

Meanwhile the burn-to-revenue gap (OpenAI reportedly spending far above revenue) pushes labs toward IPOs, trading financing necessity for public-market discipline right as the 'easy gains are ending' thesis gains plausibility. Whoever owns the customer and can fund the next training run wins; labs lacking both are the exposed layer.

Note: the specific figures (Claude Opus 4.8's ranking, Goldman's $7.5T, OpenAI's $34B/$13B) come from weak, undated sourcing and should be treated as directional, not confirmed. The structural read holds regardless of exact numbers.