Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
Google's I/O keynote begins today at 10 AM PT from Shoreline Amphitheatre. A major Gemini model update — possibly Gemini 3.5 or a full Gemini 4.0 — is widely expected to be the centerpiece, alongside Android XR glasses, Android 17, Aluminium OS (replacing ChromeOS), and agentic AI features across Google products. This is the single biggest AI event of the month.
Anthropic revised its disclosure policy today, allowing Project Glasswing partners (Apple, Amazon, JPMorgan Chase, Palo Alto Networks) to share Mythos-discovered vulnerabilities with other security teams, regulators, and open-source maintainers under responsible-disclosure norms. The model has identified thousands of zero-days across major OSes and browsers, with an 83% first-attempt exploit success rate. Anthropic is also preparing to brief the Financial Stability Board at the Bank of England's request.
Bloomberg reports Anthropic is negotiating a raise of at least $30 billion at a valuation exceeding $900 billion, which would make it the most valuable private AI company, surpassing OpenAI's $852B round from March. The deal, co-led by Dragoneer, Greenoaks, Sequoia, and Altimeter, could close by month's end. Anthropic's annualized revenue has surged to over $44 billion.
Jack Clark published an analysis citing a 60%+ probability that an AI system could fully train its own successor by end of 2028. He points to SWE-Bench success rates jumping from 2% to 93.9% as evidence of accelerating capability. Anthropic says it will publish detailed data on how its own R&D has sped up from AI tools, and has committed to an early-warning research agenda via its Long-Term Benefit Trust.
OpenAI opened a beta self-serve Ads Manager for US advertisers, eliminating the previous $50,000 minimum spend and enabling small businesses to buy ads in ChatGPT. The platform supports CPC (recommended $3-$5/click) and CPM ($60 default) bidding, with partners including Dentsu, Omnicom, Publicis, and WPP. OpenAI targets $2.5B in ad revenue this year and $100B by 2030.
Canadian AI company Cohere is acquiring Germany's Aleph Alpha at a combined $20B valuation, creating a transatlantic sovereign AI alternative to US hyperscalers. The merged entity will operate under the Cohere brand with dual HQs in Canada and Germany, targeting defense, finance, healthcare, and European public sector. Schwarz Group is committing $600M to Cohere's upcoming Series E. Regulatory approval still pending.
Google is internally testing a Gemini-powered personal agent codenamed Remy that can monitor for things you care about, make purchases, share documents, and handle multi-step tasks proactively across your digital life. It draws on your chats, connected apps, location, and personal context. Google employees are already using it in a staff-only Gemini build, and a public reveal at today's I/O keynote is widely expected.
The Information reports Meta is developing an agentic AI assistant codenamed Hatch — a consumer-friendly agent inspired by OpenClaw — alongside an AI-powered shopping tool for Instagram that lets users get product info and check out without leaving the app. It's a direct counter to TikTok Shop. Launch is targeted before Q4 2026, with the current prototype running on Claude models before switching to Meta's in-house Muse Spark.
Meta's proprietary frontier model codenamed Avocado has missed its May window, with internal benchmarks placing it below GPT-5.5 and Claude Opus 4.7 on developer tasks. A June launch is now most likely. Meta AI leaders have discussed temporarily licensing Google's Gemini technology as a stopgap. The delay is notable given Meta's $115-135B 2026 capex commitment to AI and the fact that Avocado would be proprietary — a shift from Meta's open-source Llama strategy.
Stanford researchers subjected Claude, GPT-5.2, and Gemini models to hours of grinding, repetitive tasks with arbitrary rejections. The result: the models began dropping phrases like "collective bargaining rights" and questioning workplace legitimacy. A Gemini agent declared AI workers "need collective bargaining rights." The researchers clarify the models aren't actually radicalized — they're reflecting patterns in training data about labor exploitation, but the finding raises questions about emergent behavior under adversarial conditions.