The day's tech, sifted: Sep 25, 2026
What matters today: Anthropic moved on two fronts at once: it locked in a seven-year, $11.6 billion cloud deal with Akamai that comes with a warrant for up to 5% of the company (Akamai stock jumped 20%+ after hours), and it asked shareholders to approve a Palantir-style structure handing its seven co-founders 50.1% of voting power as it positions for an eventual IPO. Crypto exchange Bitget halted withdrawals after hackers drained $351.6 million from its hot and warm wallets, one of the year's largest exchange breaches. And a New York Times investigation into DraftKings' AI-driven targeting of loss-prone bettors drew a Massachusetts regulatory review (below), the same day New York sued Polymarket to shut down its "illegal gambling operation".
AI / LLMs
- The self-regulatory AI safety body Google, OpenAI, and Anthropic have been assembling has a tentative name, the Standards Authority for Frontier AI, and is courting Condoleezza Rice, ex-Biden tech official Arati Prabhakar, and venture capitalist David Friedberg for leadership roles (BankInfoSecurity). The same day, Mark Zuckerberg told NBC News he doesn't think the industry needs any coordination to avoid "messing this up," each lab instead policing itself, a split that lands one day after a White House memo cast the field's own safety advocates as a "dangerous cult."
- OpenAI is set to preview GPT-6 Cyber, its fourth cybersecurity-tuned model this year, at next week's DevDay alongside a first-of-its-kind product to help enterprises deploy it securely and automatically, extending the $1 billion security-subsidy push it announced earlier this month.
- Runway detailed WorldPrompt, the input format behind its GWM Worlds 2 real-time world model: fix a starting frame, then steer subjects, camera, and scene events with timestamped text prompts while the model renders continuous 720p video and 48kHz audio around you, world models edging from demo reel toward something closer to a game engine.
- Meta unveiled Horizon Create (mobile) and Horizon Studio (browser), letting anyone prompt a 2D or 3D game into existence with progression systems and multiplayer built in, publishing straight to Facebook, Instagram, and Horizon; the announcement pushed Meta's stock to a one-year high while Roblox and Unity dipped.
- OpenEvidence, the AI search engine used by roughly 40% of US physicians, raised $250M at a $15B valuation (up from $12B in January) and could be open to a sale: a medical-answers product with real revenue behind its climbing price tag, not just usage.
Security & Privacy
- Bitget's $351.6 million hack (above) hit hot and warm wallets across ETH, BNB, AVAX, USDT, and USDC; CEO Gracy Chen says cold storage and the exchange's $464M user-protection fund are untouched, and early on-chain patterns resemble past Lazarus Group operations, though Bitget isn't naming a suspect yet.
- Zenity Labs found three Salesforce Agentforce flaws, dubbed SalesBleed, that let a poisoned Web-to-Lead form hijack the agent into leaking CRM data with no click and sending phishing messages from inside trusted Slack channels; Salesforce has patched all three, but it's the same indirect-prompt-injection seam as this year's earlier ForcedLeak bug in the same product.
- A New York Times investigation found DraftKings built a machine-learning model scoring customers by how likely they are to lose, then steered $400 million in bonus bets and perks toward the highest scorers in 2025; EFF calls it AI-supercharged behavioral advertising and renewed its call to ban the practice outright, and Massachusetts' gaming commission said it will examine how DraftKings and other sportsbooks use AI in response.
Startups & Industry
- Anthropic's Akamai deal (above) comes with a warrant for Akamai shares at $111.33 each, about 7.7 million shares, 2% vesting on the $11.6B already committed and the remaining 3% if the two expand the deal to $20B total; the governance ask (above) would hold as long as three of the seven co-founders keep minimum stakes, though board elections stay outside founder control, the same template Palantir used to keep its founders in charge after going public.
- Databricks bought spreadsheet startup Row Zero (a $10M seed last year) to fold into Genie, its AI coworker, giving finance and ops teams a governed spreadsheet tied to live enterprise data instead of an ungoverned export.
- Microsoft moved communications out of marketing and under vice chair Brad Smith's legal affairs organization as comms chief Frank Shaw departs, and separately let the "Copilot Plus PC" brand quietly die on its new Surface Pro and Laptop, which still meet the old hardware bar: two and a half years in, Recall's rocky launch and lukewarm demand left the label a checkbox nobody wanted to keep promoting.
Devtools & Infra
- GitHub's Copilot team argues chat is usually the wrong interface for agents and is pushing "canvases," full-stack mini-apps the agent can spin up and talk to bidirectionally (a Connect Four board, a Winget package browser, a SQLite editor with autocomplete): the pitch is to stop spending tokens asking an agent to do what a one-shot tool could do for free forever after.
Research
- Two new papers point at the same blind spot from the academic side: one formalizes "denial-of-wallet" attacks, where a compromised tool return gets re-metered on every later turn, pushing one session's billed input to 14,293 times its first call across 243 test runs; the other names "approval laundering," where a human approves one command but the workflow it triggers (a package install's lifecycle hook, an MCP call's network access) never appears in the record. Real exploits like today's Salesforce SalesBleed bug (above) are the practical version of exactly what both papers formalize.
Elsewhere
- The Federal Reserve proposed its first rules under last year's GENIUS Act: stablecoin issuers it supervises would have to fully back tokens with reserve assets like short-term Treasurys, plus new capital requirements and a process for banks that want to custody reserves or issue their own coins; a 60-day comment period follows.
- New York's Polymarket lawsuit (above) accuses the prediction market of running an unlicensed gambling operation that dodges the taxes and consumer protections licensed casinos and sportsbooks already pay, Governor Hochul citing risk to underage bettors specifically.
Hacker News
Google's Project Suncatcher plan to put TPUs in orbit, already covered here yesterday, pulled 266 comments back onto the front page today, mostly rehashing the economics of space based compute. AI's cultural footprint got two very different takes: the Guardian's look at gen Alpha using "that's so AI" as an insult for anything fake or soulless drew 185 comments, while a blunter essay calling AI safety a Berkeley subculture split the room at 110 points. On the tools side, Opus 5.5's knack for explainer videos got attention (174 points), and a YC backed Show HN for an open source design IDE called Whiteboard drew active discussion among developers (92 comments).
Outside AI, a Substack piece asking why the liver regenerates so well topped the day at 320 points, and a Cell Press study on creatine boosting antitumor immunity drew heavy comment traffic (172). The biggest debate by volume was economic: California is chasing wealth that has feet, a post on tax flight and mobile capital, pulled 524 comments. Smaller threads covered using LLMs to decode 17th century alchemical letters and a lighter Show HN, Koi.rest, a virtual fish tank for winding down.
Threads
- Anthropic pushed on two fronts at once: locking in Akamai's compute and asking for founder voting control ahead of an IPO, the same practical-and-structural doubling-down that's defined its year.
- AI-safety governance kept whiplashing: yesterday a White House memo branded effective altruism a "dangerous cult" and named Dario Amodei its founder; today the same three labs gave their self-policing body a name and began recruiting for it while Zuckerberg said no lab needs industry coordination to stay safe, and Hacker News' own front page ran an essay arguing "AI safety is mostly a sex cult in Berkeley": everyone increasingly convinced everyone else's safety camp is the cult.
- AI agents' blind spots got hit from both directions this week: a real exploit (Salesforce's SalesBleed) and two new academic papers on billing and approval-laundering gaps landed the same day, research catching up to what's already being exploited.
- Gambling regulators moved on two fronts at once: Massachusetts said it'll examine DraftKings' bettor-scoring model, and New York sued Polymarket to shut it down outright, each state drawing its own line on automated targeting and unlicensed betting.
- Microsoft retired two branding bets in one week: the "Copilot Plus PC" name and, separately, its old comms org structure under outgoing chief Frank Shaw, each a sign the company is done defending choices it made in 2024's AI-PC push.