Published today, sourced from stories within the last 24 hours.
Meta Closes Out Connect 2026: Muse Charm AI Pin and AI Game Builder
Meta wrapped its Connect keynote with a wave of announcements built around Muse, its recently launched AI agent. The headline hardware: Muse Charm, a handheld “Tamagotchi-like” AI pin that lets users talk to their Muse agents without a phone or computer, shipping this December. Meta also unveiled Horizon Create (a phone app) and Horizon Studio (a browser editor), which generate playable 2D/3D multiplayer games from plain-language prompts and publish them straight into Facebook and Instagram feeds — early access opens via waitlist next year. Markets reacted fast: Unity fell roughly 5% on the game-building news, with Roblox and AppLovin slipping about 1%, while Meta stock has rallied ~16% over the past week on its AI agent momentum.
Google Sends TPUs to Orbit with Project Suncatcher
Google announced it will launch a prototype satellite carrying its Tensor Processing Units into low Earth orbit on October 1, aboard a SpaceX Falcon 9 from Vandenberg Space Force Base. Built with Planet Labs, the refrigerator-sized craft is the first in-orbit test of Project Suncatcher — Google’s research into whether AI data centers could one day run on solar power in sun-synchronous orbit, where panels can be up to eight times more productive than on Earth. The chips passed proton-beam radiation testing at UC Davis beyond a five-year mission’s load, though skeptics note the economics only pencil out if launch costs fall below $200/kg — a level not projected until the mid-2030s.
Microsoft Merges Copilot Apps, Shifts to Usage-Based Pricing
Microsoft is consolidating its consumer and enterprise Copilot experiences — merging coding and agents into a single app — and moving to usage-based pricing, including new paid AutoPilot agents. The restructure folds what were separate developer, consumer and enterprise SKUs into one surface as Microsoft bets on agentic workflows as the next monetization layer.
Oracle Sends Force Majeure Notice on $18B Blue Owl “Project Jupiter”
AI infrastructure financing rattled markets Friday: Oracle sent a force majeure notice tied to Blue Owl’s $18 billion Project Jupiter data center project, sending Oracle shares down ~3.3% amid data center power delays. The episode underscores how power supply, transmission, cooling and permitting — not capital — have become the real bottlenecks as AI financing rounds grow. Brookings this week warned the risks of the AI investment race are shifting from corporate balance sheets into “opaque funding channels” like joint ventures, private loans and special-purpose entities.
Frontier Models Score Real Research Firsts
In a striking showcase of AI-as-researcher, OpenAI’s GPT-6 Astra broke a July 1941 German Enigma message that had resisted decipherment — reportedly writing its own Enigma simulator and Bombe along the way, per cryptographer Frode Weierud. Separately, Anthropic’s Opus 5.5 downloaded 5,000+ files from 17th-century scholar Samuel Hartlib’s papers and spotted that Hartlib and Newton used coded spellings for the same alchemical ingredient — evidence historian Benjamin Breen reads as a new finding about Newton’s sources.
Tesla Ramps Optimus Production
Sources say Tesla has ramped Optimus humanoid robot production to several hundred units per week, though the program still faces hurdles with robotic hands, automation equipment and supplier constraints, per The Information. Meanwhile, Seyond announced scaled LiDAR production targeting warehouse and sidewalk robotics — more signs the robotics supply chain is maturing.
The AI ROI Gap Widens
A fresh wave of surveys says companies are spending more on AI than ever but can’t prove it pays: Censuswide found just 31% of UK businesses using AI report positive returns; Esker found 72% of finance leaders say AI spend blew past plan; Tempo AI found 42% of tech leaders can’t see their AI spend at all. Ensono reports 61% of organizations paused or scaled back IT modernization partly due to AI budget diversion, and Gartner expects 40%+ of agentic AI projects to be killed by end of 2027. The counter-example: Bank of America says its 140 AI use cases cost $400M and delivered $800M — and it’s doubling next year’s AI budget.
Regulators Circle: State AGs Press Congress, US Banks Get Deregulatory Nod
A coalition of state attorneys general led by Minnesota AG Ellison urged congressional leaders to pass substantive AI legislation, warning in a letter that unchecked AI development threatens the financial system, critical infrastructure and national security — while pressing to preserve states’ oversight powers. On the other side of the ledger, US bank regulators rescinded prescriptive model-risk guidance and explicitly excluded novel AI from new rules, a deregulatory “Plimsoll line” moment for AI deployment in finance.
Published today, sourced from stories within the last 24 hours. Sources: The Verge, Unite.AI, Techflowpost, Progressive Robot, Chosun/The Economist via Brookings, The Information, TechRadar, CIO, Ciodive, Minnesota AG’s office, Forbes, GlobeNewswire.
Leave a Reply