▒▓█▓▒
░▒▓███████▓▒░
░▒▓███████████████▓▒░
T H E U N C E R T A I N F U T U R E S
· · all the news that fits · ·
────────────────────────────────────────
No. 6 · Wed · 17 June 2026Good Wednesday, and welcome back to The Uncertain Futures — the newsletter formerly known as Temperature Zero. The money got bigger and the questions got darker. SpaceX spent its first days as a public company buying a $60 billion AI startup, Anthropic's export fight curdled into a question about who gets to work on American AI, and DeepSeek closed the largest first round a Chinese lab has ever raised — all while a court learned that ten drones killed without anyone pulling the trigger.
░░▒▒▓▓ TOP 5 ▓▓▒▒░░SpaceX to buy Cursor maker Anysphere for $60 billion in stock, its first move as a public company
Four days after the largest IPO on record, SpaceX said on June 16 it will acquire Anysphere, the maker of the AI coding tool Cursor, in an all-stock deal worth $60 billion — exercising an option set in April that let it either buy the company outright for $60 billion in stock or pay a $10 billion break-up fee if the deal fell through. The purchase is meant to feed SpaceX's money-losing AI division, which told IPO investors it sees a $26 trillion addressable market; Cursor runs at roughly $2.6 billion in annualized business revenue — about $4 billion all in — and at one point accounted for 40 to 50 percent of Anthropic's revenue, the supplier whose Claude Code it now competes with. The deal helped lift SpaceX past Amazon to the fourth-most-valuable US company, and lands the week Anthropic took its own frontier models dark. Covered by TechCrunch, Business Insider, Axios, GeekWire, The Information, and NZZ. Primary: CNBC · TechCrunch
Washington's Anthropic ban hardens into a fight over foreign AI talent as Monday's truce talks fail (update)
The export order that took Fable 5 and Mythos 5 dark, as we previously covered, has widened: the directive bars foreign nationals — including Anthropic's own non-citizen engineers — and rival labs now read it as a precedent rather than a one-off. On Saturday OpenAI strategy chief Jason Kwon told staff the company had warned the administration that building AI "requires the best talent from around the world," and The Information reports OpenAI and others are alarmed the White House has put a target on the foreign-born researchers who fill US labs. Anthropic flew its Frontier Red Team to the Commerce Department's AI-standards office on Monday to walk officials through the model's safeguards, but the meeting yielded no resolution; both sides say they are working to fix it. Deutsche Bank's Jim Reid, on Anthropic's IPO risk, told clients the lesson for buyers is plain — you can't rely on something that could be switched off. Covered by The Information, Axios, Bloomberg, POLITICO, WIRED, and The Guardian. Primary: The Information · Reuters
DeepSeek closes a record $7.4 billion first raise, structured so founder Liang Wenfeng keeps absolute control
DeepSeek — the Chinese lab whose cut-rate models have been driving the AI price war that now has American labs cutting prices too — closed its first outside funding round at more than 50 billion yuan ($7.4 billion), valuing it above $50 billion, The Information reported on June 16. Founder Liang Wenfeng, the quant who built the High-Flyer hedge fund and spun DeepSeek out of its compute in 2023, structured the deal to keep absolute control: investors must park their money in a limited partnership he manages personally rather than in DeepSeek itself, accept a five-year lockup, and forgo voting rights. Liang is writing the biggest check at 20 billion yuan, ahead of Tencent's 10 billion and battery maker CATL's 5 billion, with JD.com, NetEase, and IDG Capital each adding 3 billion; only China's national AI fund, putting in 1 billion yuan directly, escapes the lockup and keeps a vote. The raise lands as DeepSeek's commercial pull continues to grow: Microsoft — OpenAI's own biggest backer — said this week it is weighing a self-hosted DeepSeek V4 as a cheaper, optional model inside Copilot Cowork, the enterprise agent that runs on Anthropic's and OpenAI's models now, as agent costs balloon. Covered by TechCrunch and The Next Web. Primary: The Next Web · Axios
ChatGPT's share of the AI-assistant market falls below half for the first time, to 46.4 percent
ChatGPT slipped under 50 percent of the global AI-assistant market for the first time, to 46.4 percent, per Sensor Tower's 2026 report — still the clear leader, still 1.1 billion monthly users, but no longer the default the whole market bends around. Google's Gemini reached 662 million users and 27.7 percent, Anthropic's Claude 245 million and 10.3 percent, as enterprises spread their bets across models and cheaper Chinese challengers undercut everyone on price. It reads less as a collapse than a normalization — the first real crack in a market OpenAI once had to itself. Covered by TechCrunch and Sensor Tower. Primary: TechCrunch
Google sues a China-based ring that used Gemini to mass-produce the scam texts flooding US phones
Google filed suit on June 12 against Outsider Enterprise, a China-based cybercrime network it accuses of using its own Gemini model to generate phishing pages and blast millions of fraudulent texts — the first time Google has sued anyone specifically for misusing Gemini to run scams. In two weeks in May the group sent some 2.5 million scam texts to Android users; the FBI ties it to 3.87 million stolen card numbers and about $1.9 billion in losses since July 2023. The operation ran like a franchise: an $88-a-week kit on Telegram with 290-plus templates impersonating Google, the US Postal Service, and banks, turning what once took a coder and a few hours into a job for anyone with a Telegram account. Covered by The Neuron, Future Blueprint, and Help Net Security. Primary: Help Net Security
░░▒▒▓▓ STATS OF THE DAY ▓▓▒▒░░░░▒▒▓▓ AI ▓▓▒▒░░░░▒▒▓▓ TECH ▓▓▒▒░░░░▒▒▓▓ POLICY ▓▓▒▒░░░░▒▒▓▓ BUSINESS ▓▓▒▒░░░░▒▒▓▓ GEOPOLITICS ▓▓▒▒░░░░▒▒▓▓ ALPS ▓▓▒▒░░░░▒▒▓▓ RESEARCH ▓▓▒▒░░░░▒▒▓▓ COMMS DESK ▓▓▒▒░░KPMG pulled an October report on enterprise AI adoption after the Financial Times found it cited fabricated case studies, including made-up gains at UBS and Swiss Federal Railways; the firm said it had removed the report and expects staff to verify AI output against independent sources (SWI swissinfo.ch). Takeaway: an AI-assisted draft that invents a flattering client win is a reputational landmine with your own logo on it — the verification step you skip to ship faster is the one a reporter will run for you.
░░▒▒▓▓ ONE MORE THING ▓▓▒▒░░FIFA's sponsorship rules forced Levi's to drape a white tarp over its own name at Levi's Stadium in Santa Clara, the venue it pays to put its name on, lest a non-sponsor logo sully a World Cup match. Rather than sulk, the denim brand played along and covered the logo on its Instagram profile with a matching white tarp — a company censoring itself, gamely, on its own turf (Mashable).
░░▒▒▓▓ CORRECTION ▓▓▒▒░░Yesterday's edition reported the US-Iran deal as already signed. It wasn't: a framework was agreed at the G7, but no document is signed until Friday's ceremony at Bürgenstock (see Geopolitics above).
░░▒▒▓▓ TRACK OF THE DAY ▓▓▒▒░░🎵 Serpentine 4 — OK EG, Priori
Back tomorrow, conditions permitting — your editors at the Uncertain Futures