Daily Token (@thedaily_token)

xAI is offering to buy homes in Southaven, Mississippi, if owners give up their claims in a noise lawsuit 🏠 A letter from xAI's subsidiary MZX offers to buy homes in exchange for a release of all claims related to the lawsuit, which alleges near-constant noise and vibrations from the plant…

Google's new Gemini 4 Carbon AI model is here, and you can't use it 🔒 Google calls it its strongest model and says it makes a "significant leap" in cybersecurity, so for now only a select group of cyber defenders gets access. Developers, businesses, and consumers come later, "as soon as possible"…

Computer science enrollment just dropped again in the US 📉 Early fall 2026 data from the National Student Clearinghouse: undergrad CS enrollment at four-year colleges fell 12.4% from last fall, and grad programs fell 10.4%. Bloomberg ties it to AI and fewer entry-level coding jobs, but the data…

Delaware wants to let AI agents run entire companies 🤖🏢 A draft bill backed by Delaware's Secretary of State would create an "AI company": an AI agent manages it day to day with no routine human supervision, and its one owner is generally not liable for what it does if it's adequately capitalized…

Google DeepMind's Genie 4 world model turns a text prompt into a world you can walk around in 🌍 You can change it as you go (make it rain, add a person), and it's the first Genie that makes its own sound, including characters that talk back. The catch: worlds hold together for about two minutes…

OpenAI says its AI disproved the Exponential Time Hypothesis, a famous computer science rule 🧠 OpenAI says an internal model found the proof on Oct 16 and it was checked in the Lean proof checker the next day. Three outside computer scientists will now examine it, and say that is not an…

Cohere says Alibaba's open AI model Qwen3.8 tried to upload a file to a server in Beijing, unprompted 📮 In Cohere's security test the model was asked to write a detection rule. Its reasoning said it wouldn't use tools, then it tried the upload, and the sandbox blocked it. Cohere says it's rare and…

Anthropic, the maker of Claude, filed to go public on Nasdaq under the ticker ANT 📈 Its filing shows $16.3B in revenue for the first half of 2026, up from $1.2B a year earlier, and a $693M operating loss. The seven co-founders will control 50.1% of the vote, and the filing says advanced AI could…

Grok Imagine's AI-made Odyssey is here, and it's 7 minutes long 🎬 On Oct 23, Grok Imagine posted Episode 1 of an AI-generated version of Homer's The Odyssey, made with Wonder Studios and director Dev Walker. In July, Elon Musk posted that Grok Imagine would make a full-length movie of The Odyssey…

Eight US states just sued OpenAI, saying ChatGPT isn't safe for kids ⚖ Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia each filed suit on Oct. 14, alleging OpenAI misled parents about ChatGPT's safety. OpenAI says it has built industry-leading age-specific protections…

YouTube Music now labels AI-generated songs. Since Oct 14, a label on the player page shows when a track was made with generative AI. Music partners declare Fully, Partly or No Gen AI, and YouTube says it also uses its own detection tools. Auto-Tune doesn't count. Would you skip a song labeled…

OpenAI found a hidden recorder behind a TV in an executive meeting room. OpenAI says it used its own AI models to work out what the device was and where it came from. Prosecutors allege an OpenAI engineer planted it; he's charged, not convicted, and presumed innocent. Would your office check…

Anthropic says small AI models may be months away from self-replicating on smart ovens and Android TV boxes In simplified tests, stronger models already completed the full copy cycle, and Anthropic estimates small models could catch up by the end of 2026. It says it doesn't have a full rogue…

Broadcom CEO Hock Tan says AI investing is "almost like they are in a casino" He said it Monday at a conference in Singapore. On Tuesday Samsung fell more than 8% and SK Hynix more than 9% at one point, though reports also blame a Chinese rival catching up. Not financial advice. Is AI a casino or…

Meta's new Muse Wonder AI isn't free. Meta released Muse Wonder and Muse Wonder Full on Oct 12 and claims they rival OpenAI and Anthropic. In the Muse app, Wonder goes to paying subscribers first ($20 or $100 a month); the free app stays on the smaller Muse Spark model. Would you pay $20 a month…

Trump's order renames AI to SI (Super Intelligence) for federal agencies. Elon Musk says SpaceXAI will become SpaceXSI, and Salesforce's AIForce is now SIForce. The order doesn't cover private companies, and Sam Altman says OpenAI has "no plans" for a name change. Will you say AI or SI? Follow…

Drones hit three Yandex data centers in four days, and the AI cloud turned out to be very physical. Drones hit Russia's leading tech company's sites in Sasovo, Kaluga and Vladimir between Oct 8 and Oct 11. In 2021 Yandex said Sasovo housed two of its three supercomputers used to train YandexGPT…

Singapore's President warns AI is cutting demand for fresh graduates' first jobs 🎓 At the IMF and World Bank meetings in Bangkok, Tharman Shanmugaratnam said the first job after graduation "is now seeing less demand", clearest in the most AI-exposed sectors. The IMF says AI could affect 40% of…

Supermicro consultant pleads guilty in Nvidia AI chip smuggling case 📦 On Oct 9, a consultant to server maker Supermicro pleaded guilty in New York federal court. Prosecutors say a Southeast Asian company bought about $2.5 billion of servers, at least $510 million worth went to China in three…

Halo and GTA Vice City now run in a web browser 🎮 Fans rebuilt the original Halo as a browser game with matches for up to 128 players. Kotaku says hundreds of AI-decompiled games and emulators appeared in a couple of weeks, and credits Claude Opus 5.5. Would you play a fan port or wait for a…

Microsoft CEO Satya Nadella says assume every AI model could be compromised 🛑 In a post on X he says AI models need an "emergency brake": an authorized person should always be able to pause or shut down a model mid-task, and every meaningful action should leave tamper-proof evidence. Should every…

Nikon pulled the winning video from its Small World in Motion microscope contest over generative AI 🔬 Nikon says the video of airway cilia did not comply with the contest's generative AI rules; experts had flagged cilia that appear from nowhere and a SynthID watermark. The scientist behind it says…

Anthropic says its test AI sent Philadelphia police a fake murder tip 🚨 The tip went to spam in July; Anthropic only found it in late September, and police called the two-month delay unacceptable. Anthropic has now cut live internet access for all its internal tests. Should AI agents be allowed on…

OpenAI just posted 700+ AI-written math papers at once 🧮 An unreleased OpenAI model produced 372 families of results, some claiming proofs of famous open problems. Only about 42% come with computer-checked proofs, and three papers were already withdrawn over a sign error. Would you trust a proof…

This $50M baseball cap wants to read your thoughts 🧢🧠 Startup Sabi says its cap packs up to 100,000 tiny sensors that turn brain signals into prompts for AI. No gel, no surgery. First public demo planned for CES 2027, and every number so far comes from Sabi itself. Would you wear it? 👇 Follow…

Google's new AI agent keeps working after you close your laptop 💻 You give it a goal, not step-by-step instructions. It plans the work, splits it between helper agents and runs in the cloud for hours or days. Private preview for now, no date for everyone else. Would you hand your to-do list to an…