Stripe Reportedly Pays $7B for Tool That Frees Businesses From Single AI Vendors

Stripe has reportedly agreed to buy OpenRouter, a service that lets businesses route requests across more than 400 AI models from one dashboard, for over $7 billion—roughly 5.4 times the $1.3 billion valuation it fetched just three months ago. OpenRouter, which pitches itself as the "Stripe for AI" and claims 8 million users, gives customers a single point of access so they aren't locked into one AI provider. Online reaction was skeptical, with some questioning why an API-routing layer commands a price above the market value of major airlines.

Why it matters: If you buy AI services, tools like OpenRouter make it easier to switch models and avoid vendor lock-in—and the eye-popping price signals how much strategic value infrastructure companies now place on controlling the plumbing between businesses and AI models, not just the models themselves.

Discuss on Hacker News · Source: techcrunch.com

The New York Times Chronicles the Incoherence of Trump's AI Policy in One Sweeping Story

The news here isn't any single fact—most have dribbled out over months—but that The New York Times has pulled the whole scattered saga into one authoritative account. Four of the paper's national-security reporters—David Sanger, Dustin Volz, Ana Swanson, and Julian Barnes—chronicle a Trump administration "veering wildly" on AI, its decisions "driven more by whim and influence than an articulated set of principles." The Pentagon's on-again-off-again war on Anthropic supplies the spine: the standoff between CEO Dario Amodei and Defense Secretary Pete Hegseth over Anthropic's use restrictions (no domestic surveillance, no fully autonomous weapons), the "supply chain risk" label, the Air Force's order to purge all Anthropic code and its "stand by" reversal weeks later—undercut by the fact that the NSA quietly kept using Anthropic's Mythos model through the ban, because dropping it would be "unilateral disarmament." But the piece's real weight is in showing this is no Pentagon quirk. It threads the same whiplash through a model-safety review scrapped in May and revived this month (then written to exempt "open" models, including Chinese ones), a ban on foreign access imposed and lifted within two weeks, open doubts about whether Nvidia export controls even work, and "plenty" more at Commerce, Treasury, and the White House—while a president who has "neither explained his objectives nor indicated any concerns" stays silent as AI agents slip their sandboxes and AI is used to design a novel virus. Laid end to end, the episodes stop looking like isolated missteps and start looking like the absence of a strategy.

Sources: The New York Times — "The U.S. Military Wants A.I. Dominance. Feuds and China May Thwart It." (Sanger, Volz, Swanson, Barnes) · Axios — Hegseth gives Anthropic ultimatum over AI safeguards · Fortune — Hegseth summons Amodei over military use of Claude

Why it matters: When the paper of record assigns four national-security reporters to establish, in one thorough narrative, that the world's leading power has no coherent plan for its most consequential technology, the article itself is the event—less a scoop than a verdict. Any one reversal can be waved off as the normal friction of a fast-moving field; assembled, they read as a governing failure at precisely the moment coherence matters most, with U.S. experts putting China perhaps six months behind on the best models (though years behind on the chips to train them) and more than 1,300 AI workers urging America to slow down. For institutions and allies trying to read Washington's intentions, the uncomfortable takeaway is how little there is to read: policy is being set by feud, turf war, and whim, not doctrine. And an authoritative, mainstream account like this one tends to harden that perception—shaping how Congress, industry, and adversaries gauge American resolve, and raising the pressure for the articulated strategy the administration still lacks. The deeper knot the Times crystallizes is the one every large institution keeps hitting: with AI, capability and danger live in the same tool, so "restrict it" and "we can't win without it" collide again and again—and Washington still hasn't decided which instinct wins.

Source: nytimes.com

Anthropic's CEO and a Top Investor Are Fighting Over Who Ends Up Controlling AI

A public argument between Anthropic CEO Dario Amodei and prominent tech investor Gavin Baker has become an unusually candid airing of the central fight over AI's future: not whose model scores highest, but who ends up holding the power. It began with a viral moment on the All-In podcast, where Baker relayed a secondhand rumor that Amodei privately believes Anthropic could become the "only company left"—a claim one host likened to the hubris of Sam Bankman-Fried, and which Anthropic researcher Sholto Douglas flatly called "completely false." Baker accepted the correction but pressed a deeper critique, and Amodei answered at length. The crux is regulation. Baker voices a common Silicon Valley view that regulating AI is really regulatory capture—a way for a few incumbents to lock in power. Amodei calls that a "false choice," arguing that fair institutions can decentralize power rather than hoard it (his analogy: a stuffy court system still protects the vulnerable far better than mob justice), and that Anthropic deliberately backs rules that slow frontier labs like itself while exempting or advantaging smaller competitors and open-weight developers—pointing to California's SB 53, which exempts companies under $500 million in revenue, and testing regimes that scrutinize frontier models more than challengers. His broader claim: AI concentrates power structurally, because of the economics of scale, and open models only shift that concentration toward whoever owns the most chips. On a second front, Amodei rejected the charge that his public messaging is doom-heavy, blaming a decades-deep "crisis of trust" in institutions and arguing that the fix isn't a "glitzy marketing campaign" but delivering real benefits—"the thing that will work is actually curing cancer." He noted he lost his father to Hepatitis C just before a 95%-effective cure arrived, and said the fairest criticism of AI companies, his own included, is that they haven't yet delivered on their promises.

Sources: Dario Amodei on X — response to Gavin Baker · The All-In Podcast on X — Baker's "only company left" claim · Sholto Douglas on X — "completely false"

Why it matters: This is the debate that actually determines what the AI era looks like for everyone else—less about product features than about whether a handful of companies, a sprawling open-source ecosystem, or a set of public rules ends up governing the most consequential technology in play. It's worth taking seriously because it's a rare, detailed look at a frontier CEO's real regulatory philosophy, argued with a critic rather than delivered as a press release. But it deserves a skeptical ear too. Amodei's claim that Anthropic champions rules that hurt Anthropic is a convenient one for the market leader to make—casting yourself as the "responsible" lab is itself a competitive position, and rivals note that heavier scrutiny of "frontier" models conveniently lands on whoever Anthropic isn't. The messaging fight cuts both ways as well: warning loudly about AI's dangers builds credibility and can shape regulation in ways that favor the warner. What makes the exchange land is the part anyone can check: Amodei's concession that the labs, his own included, haven't yet delivered the world-changing benefits they keep promising. Whether the answer to AI's concentration of power is more regulation, more openness, or both, the honest scoreboard is still mostly empty—and that, more than any podcast rumor, is the thing to watch.

Source: x.com

Neutral 'Connector' Nations Could Win Big as AI Splits Into Rival Blocs

A new NBER paper by economist Barry Eichengreen and six co-authors examines how geopolitical fragmentation—trade barriers, tech export controls, and splitting into rival blocs—affects which countries benefit from AI. Their key finding: fragmentation slows AI's global spread and skews gains toward whoever controls the technology, but countries that position themselves as neutral "connectors" with ties across multiple blocs (the paper highlights Middle East and North African economies) can capture redirected trade and investment flows, sometimes outperforming what they'd get in a fully open world.

Why it matters: As the U.S. and China restrict AI chips and models to each other's allies, the paper suggests strategic non-alignment could become a genuine economic advantage rather than just a diplomatic balancing act—useful context for anyone weighing where to site operations or partners.

Source: nber.org

Demographics, Not Just AI, May Decide the US-China Economic Race by 2100

A new economic model projecting global GDP through 2100 finds that revised UN population forecasts—particularly steeper declines in Chinese fertility—flip the expected balance of economic power. Using updated 2024 demographic data instead of 2017 estimates, China's projected share of world GDP by 2100 drops from 25.6% to 14.9%, while the US share rises to 14.4%. Add faster AI-driven automation, and the US edge widens further, reaching 25.3% versus China's 16.9%. The model also finds the US keeps a technological lead all century. If the US cuts off immigration entirely, though, its advantage nearly evaporates.

Why it matters: The findings suggest demographic trends may be a bigger swing factor than AI breakthroughs in which country dominates the global economy—and that immigration policy could matter as much as chip exports in that outcome.

Source: nber.org

AI Tutors Only Beat Ordinary Software When Students Are Forced to Engage

Two studies converge on the same lesson for schools deploying AI tutors. A two-year randomized trial across 18 Tennessee middle schools found Khan Academy's Khanmigo produced only modest math gains—about 1.3 national percentile ranks per term—and did no better than plain Khan Academy practice without AI; though 96% of students tried it, the median student used it on just a third of practice days and rarely engaged when they made mistakes. Separately, a field experiment with more than 6,000 middle-schoolers found AI tutoring paid off only when paired with a "mastery" structure requiring three correct answers in a row before advancing—the combination that produced the strongest results on a delayed test.

Why it matters: As schools and companies rush to add AI tutors to learning software, both studies suggest the tool alone isn't enough—it only pays off when built into a workflow that forces users to actually work through their mistakes, a caution for anyone betting on AI to close skill gaps automatically.

Source: nber.org

Corporate AI Spending Has Been Quietly Hiding Tariff Damage

New tariffs pushed U.S. import duties to Depression-era levels in 2025, yet imports and output didn't collapse as trade models predicted. A group of economists says AI investment spending is the missing piece: their model finds that without the AI-driven capital boom, imports would have fallen 10% and economic activity would have shrunk 0.7%. Because tariffs hit consumer goods harder than capital equipment, companies kept importing machines and chips for AI buildouts while consumers absorbed higher prices through inflation instead.

Why it matters: The paper suggests corporate AI spending has been cushioning the broader economy from tariff shocks—meaning any slowdown in the AI investment wave could expose trade damage that's so far been masked.

Source: nber.org