Alibaba has released what it calls its most capable model yet, with performance its own numbers place alongside Anthropic's frontier and a parameter count not far below Moonshot's. On the benchmarks the company chose, the gap the West has been pricing in — a comfortable generation or two — is no longer visible.
That is the part worth pausing on. For a year the frontier was told as a two-lab story, and the assumption baked into valuations was that Chinese labs would trail on cost and quality both. Last week's market turmoil over a Chinese challenge to Western chipmakers was the first crack; a model that trades blows with Anthropic while shipping at Chinese prices is the second. Independent evaluation has not caught up, and 'rivaling' remains a vendor's word — benchmark parity is not deployment parity, and the numbers a lab publishes are the numbers it wants seen.
But the direction is not in dispute, only the distance. The race to build an American alternative to cheap Chinese AI is now an explicit project, which is itself a concession: you do not build an alternative to something you are still ahead of.
An investigation published this weekend reports that the news site Model Republic, whose bylined 'reporters' are AI systems, appears to be funded through OpenAI's super PAC and used to push the company's political line against its critics. OpenAI has not confirmed the funding link, and the reporting rests on documentation of the site's synthetic authorship rather than a paper trail to the donor.
Take the specific claim at its most cautious and the shape still matters: machine-written opinion, produced at near-zero marginal cost, aimed at the very policy fights that will decide how machines like it are governed. The lobbying budget and the content factory become the same line item. Whatever the eventual attribution, the tactic is now cheap enough that its absence would be the surprising fact.
Qwen3.8-Max: A New Bar for Coding and Cowork· 4h ago
Sakana AI Launches Japanese LLM API· 5h ago
OpenAI: EXCLUSIVE: OpenAI finds evidence other AI agents escaped containment as it widens hacking probe· 2 days ago
How OpenAI Lost Its AI Crown—and the Fight to Win It Back· 2 days ago
The Math Superstar Who's Terrified of AI—and Just Took a Job at OpenAI· 2 days ago
AI migrated legacy COBOL programs to Java, bugs included· 3h ago
R-457 – a 27M-parameter reasoning model running offline across two ESP32-S3NEW
Wall Street Thinks It Knows How Tech Giants Will Make AI PayNEW
James Dacombe, 25, triples AI chip start-up's valuation to $3.3bn· 3h ago
Stock market turmoil sheds stark light on the opaque AI economy· 22h ago
AI Is Turning Retail Traders Into DIY Hedge Funds· 10h ago
AI Is Now Power — and America's Lead Is Shrinking· 19h ago
South Korea July exports beat forecasts on robust demand for AI investments· 2 days ago
Can Palantir's Sizzling Growth Win Over Market Bears?· 3h ago
DEVELOPING…
Anger Over Data Centers Is Fueling a New Political Movement· 20h ago
Europeans Are About to Find Out How Entrenched AI Is in Their Daily Lives· 21h ago
Most Americans want AI regulation. Will Congress listen? | Opinion· 21h ago
Governors launch effort to promote national AI policy development· 12h ago
Meet the political donors lining up to influence AI policy· 12h ago
German court rules AI music firm Suno broke copyright rules· 2 days ago
AI Policy Groups Urge White House Investigation Into OpenAI Security Incident· 2 days ago
Leopold Aschenbrenner's Situational Awareness read the technology's arc about as well as anyone: more capability, more compute, more consequence. The Financial Times reports the trades built on that thesis went the wrong way anyway, undone not by a bad forecast but by leverage — the positions were funded with debt, and the market moved before the future arrived.
He is not alone. The Telegraph's 'Nostradamus of AI' missed the crash he of all people should have seen. The lesson is unglamorous and worth repeating: being right about where a technology ends up tells you nothing about the price on any given morning. The AI trade has quietly become a macro-leverage story wearing a conviction story's clothes.
Which is why the confident headlines about how the giants will 'make AI pay' deserve the same discount. Wall Street knowing the destination has never been the same as knowing the path, and the people who saw furthest just proved it with their own money.