• Cursor AI raised another $2.3 billion and launched its in-house coding model, "Composer", which is an abbreviation for "Creating Our Master Plan: Obliterate Software Engineers, Reluctantly". — CNBC

• Anthropic announced a $50 billion investment to build custom AI data centers in Texas, cause nothing cools a GPU cluster better like 113°F shade. — Anthropic

• Thinking Machines Lab, the new startup from ex-OpenAI CTO Mira Murati, is at a $50 billion valuation before there's even a product yet, it's like when an Avenger goes solo and the box office still hits a billion. — Bloomberg

• SoftBank sold its entire $5.8 billion stake in Nvidia to fund a $30 billion investment into OpenAI and its infra. It’s like trading a fully-evolved Charizard for a Mewtwo egg someone claims will hatch ‘soon.’ — TechCrunch

• Salesforce acquired agentic AI startup Doti AI for a reported $100 million. I think Marc Benioff just looks at AI startups like: ‘Yeah, put it on the company card.’ — Calcalistech

• Microsoft is investing $10B in a new AI hub in Portugal, officially making the country’s GDP: tourism, sardines, and ChatGPT.” — Bloomberg

• Alembic Technologies raised $145M for causal AI to explain why things break. Rumor has it the system prompt is just: “Say DNS” — SiliconANGLE

• Wonderful, raised $100M to build AI customer service agents, finally giving companies a way to ignore customers at scale — TechCrunch

• OpenAI released GPT-5.1 with Instant mode, Thinking mode, and personality controls, and now AI has more moods, and emotional range than I’ve had in my entire adult life. — OpenAI Blog

• Google DeepMind dropped SIMA 2, an AI that learns and reasons in 3D worlds, perfect for anyone who thought their NPCs weren’t smug enough already — deepmind.google

• World Labs, launched Marble, an AI that builds editable 3D worlds from prompts, giving every user their own personal game engine without the emotional damage of learning Unity. — worldlabs.ai

• Google launched “agentic checkout,” an AI that buys things automatically, because impulse purchases clearly weren’t happening fast enough. — Google Blog

• Google rolled out six new AI features for its Photos app, now you can edit photos with Nano Banana, which is a very clever way of turning “remove background” into “your storage is full.” well played Google, well played — Google Blog

• Baidu unveiled its next-gen flagship model “ERNIE 5” powered by its two in-house AI chips. Nvidia was not amused. — VentureBeat

• Anthropic makes Claude Code available in the browser and iOS, finally letting you write code in the same place you Googled how to write code — X (@claudeai)

• Meta integrated Meta AI to Facebook Marketplace, which is perfect because what that platform really needed was smarter scammers — TechCrunch

• ElevenLabs dropped 'Scribe v2 Realtime' and launched a Voice Marketplace for celebrity AI voices, perfect for anyone who wants their grocery list transcribed by Morgan Freeman in real time — elevenlabs.io

• OpenAI began piloting ChatGPT 'Group Chats' allowing up to 20 people in a shared conversation. Finally, a group chat where the AI is the only one actually reading the messages. — OpenAI Blog

• LinkedIn rolled out an AI-powered search feature that allows users to find people using natural language queries. Huge news for anyone who’s spent a decade trying to find someone without knowing 14 exact filters and their job title from 2015. — LinkedIn News

• Mozilla introduced an opt-in "AI Window" in Firefox, allowing users to interact with an AI assistant while browsing. Expect the first community response to be a patch titled “remove-ai-window.diff”. — Mozilla Blog

• AI Studio (Epidemic Sound): Generates video soundtracks. Music Business Worldwide

• Parallel Search: New search API for agents: Finally your AI can stop pretending it “doesn’t know that” and go look it up like a normal adult.. parallel.ai

• Auth0 for AI Agents: this makes sure an agent is who it claims to be, Because the last thing we need is some rogue bot logging into production and “accidentally” deleting your weekend. auth0.com

• Google AI Maps Builder: AI to build Maps or something, and now anyone can build custom maps, including your uncle who still gets lost in IKEA. Absolute power in the wrong hands.. TechCrunch

• Google Code Wiki: nice, this automatically updates code documentation, Great news for developers who treat docs like a haunted attic: never touched, always outdated, and everyone is scared to open it. Google Developers Blog

• Waymo expanded its autonomous Robotaxi services to include freeway routes in San Francisco, Los Angeles, and Phoenix. If this AI can handle LA traffic, we've hit AGI. — waymo.com

• Yann LeCun, Meta's Chief AI Scientist, is leaving Meta to launch a world-model startup, a classic case of a side quest getting so big it replaces the main storyline. — Financial Times

• Breaking Rust, a fully AI-generated country singer, just hit No. 1 on Billboard, finally giving Nashville what it’s always needed: a singer who can’t get canceled — Newsweek

• Intel's CTO, Sachin Katti, has left the company to join OpenAI, which is basically the industry version of leaving the Titanic before it hits the iceberg, but you didn't hear it from me — CRN

• Disney+ is exploring AI tools for user-generated clips, because nothing says “family friendly” more like letting the internet generate whatever nightmare it wants with Mickey Mouse in it — The Hollywood Reporter

• AIdol, Russia’s first humanoid robot, fell on stage during its debut, sources say it was calibration issues, which we all know its a code word for too much Vodka for a weekday — Newsweek

• Anthropic stopped a large Chinese hacking operation after they tried to weaponize Claude Code, proof that even state-sponsored espionage has devolved into “hey AI, do crimes for me plz.” — Anthropic

• OpenAI was found to have violated copyright law by a German court after ChatGPT was shown to have trained on protected song lyrics without permission, apparently Germany spotted a whiff of “99 Luftballons” and immediately said "nein!" — The Guardian

• OpenAI is fighting a court order demanding 20 million user chats for the NYT, because nothing says “strong journalism” like asking a judge to hand over everyone’s private prompts. — OpenAI Blog

• Anthropic introduced 'Paired Prompts' to evaluate political neutrality in models, which is the AI equivalent of trying to defuse a bomb with both teams shouting instructions — Anthropic