Tech News — 2026-04-19#

Story of the Day#

Google’s Aletheia AI has crossed a massive threshold in automated reasoning, successfully solving 6 out of 10 novel, research-level mathematical problems without any human intervention. Powered by the Gemini 3 Deep Think architecture, this marks a fundamental shift from AI acting as a generative assistant to operating as a rigorous, autonomous researcher capable of self-filtering flawed answers.

Top Stories#

The NSA is reportedly using Anthropic’s new model Mythos · Engadget The NSA is secretly using Anthropic’s new “Mythos Preview” model for cybersecurity tasks, directly defying a February presidential order that banned the company’s services from government agencies over military safeguard disputes. The fact that roughly 40 organizations currently have access to the model exposes the messy reality of federal AI procurement, as defense and intelligence agencies quietly bypass political directives to secure access to cutting-edge tools.

Blue Origin’s rocket reuse achievement marred by upper stage failure · Ars Technica Blue Origin finally managed to successfully recover and re-use a New Glenn first-stage booster, but the mission was overshadowed by a catastrophic upper-stage failure. The botched second stage delivered AST SpaceMobile’s cell-tower satellite into a uselessly low orbit, delivering a critical blow to Jeff Bezos’s heavy-launch ambitions just as the company attempts to challenge SpaceX for NASA Artemis lunar contracts.

Cloud development platform Vercel was hacked · The Verge Vercel, a dominant cloud platform for web developers, suffered a significant security breach stemming from a compromised third-party AI tool. The hacking group ShinyHunters is now attempting to sell stolen customer data, including employee names, emails, and activity timestamps. The incident is a glaring reminder of the cascading supply-chain risks introduced by plugging external AI tools into core development pipelines.

Tesla is rolling out its Robotaxi service to Dallas and Houston · Engadget Tesla is aggressively expanding its Robotaxi service into Dallas and Houston, encroaching directly on territories already occupied by Waymo. While Tesla’s marketing pushes the narrative of fully autonomous, unsupervised vehicles, the company has previously admitted that some of its Robotaxis rely on remote human operators—underscoring the blurry line between true autonomous transit and mechanical parlor tricks.

Palantir posted a manifesto that reads like the ramblings of a comic book villain · Engadget Palantir CEO Alex Karp published a bizarre 1,000-word manifesto on X to promote his new book, calling for universal national service, the “undoing” of post-WWII pacifism in Germany and Japan, and the aggressive development of AI weapons. It’s a stark, unapologetic mask-off moment that explicitly ties Silicon Valley’s engineering elite to a hyper-militarized vision of Western hard power, while simultaneously denouncing corporate inclusivity and “regressive” cultures.

Also Worth Knowing#

• The next Mac Studio and MacBook Pro releases could be postponed by several months: Apple’s next-generation Mac Studio and touchscreen MacBook Pro have been delayed—pushing the Studio to October at the earliest—due to a severe global RAM shortage exacerbated by the hardware demands of local AI models.
• Beijing’s robot half-marathon is back for its second year with far less embarassing results: At the Beijing robot half-marathon, Honor’s autonomous bipedal robot “Lightning” destroyed the human world record with a 50-minute finish time, a massive leap from the comical, remote-controlled stumbling seen at last year’s event.
• Crypto Hack Worth $290 Million Triggers DeFi Contagion Shock: A devastating $290 million cross-chain bridge exploit over the weekend has sent shockwaves through the crypto sector, draining vital infrastructure and triggering widespread contagion across decentralized finance platforms.
• Duolingo CEO Says They’ve Stopped Tracking Employees’ AI Use for Performance Reviews: Duolingo CEO Luis von Ahn reversed a controversial policy that tracked employees’ AI usage, admitting that forcing “AI for AI’s sake” was unfairly penalizing workers whose jobs weren’t genuinely improved by the technology.
• Disney Creates Its Own IMAX for ‘Avengers: Doomsday’ After Losing Screens to ‘Dune: Part 3’: Blocked from dominating real IMAX screens this holiday season by Dune: Part 3, Disney has simply invented its own premium large format certification dubbed “Infinity Vision” so it can still charge audiences a massive premium for the upcoming Avengers: Doomsday.