Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
AI chipmaker Cerebras went public on May 14 in the biggest U.S. tech IPO since Uber, raising $5.55 billion. Shares opened at $350 (from a $185 IPO price) and closed up 68%, giving the wafer-scale chip company a market cap approaching $95-106 billion. Demand was more than 20x oversubscribed.
Google's annual I/O keynote on May 19 will debut Gemini 3.2 Flash, designed to rival OpenAI's GPT-5.5 at roughly 92% of its performance while slashing inference costs to 1/15th–1/20th the price. Query latencies target sub-200ms. Google is also expected to announce "Gemini Intelligence" as an agentic OS layer for Android.
Anthropic is in discussions to raise $30–50 billion at a valuation up to $950 billion, which would surpass OpenAI's $825B for the first time. The company reported 80x year-over-year revenue growth in Q1 2026 and has secured compute commitments from Google ($40B) and Amazon ($25B).
In a stunning reversal, Elon Musk's SpaceX is handing Anthropic the entire capacity of its Colossus 1 data center in Memphis — over 300 megawatts and 220,000 Nvidia GPUs. xAI is being dissolved as a separate company (now "SpaceXAI"), and Musk said he was "impressed" after spending a week with Anthropic's team, despite calling them "evil" just three months ago.
OpenAI shipped Codex access inside the ChatGPT mobile app for iOS and Android, letting developers approve tasks, review diffs, and monitor terminal output from their phones. More than 4 million people now use Codex weekly. The mobile app acts as a remote control for Codex running on Mac (Windows support coming later).
Anthropic launched "Dreaming," a scheduled process that reviews an agent's past sessions, extracts cross-session patterns, and curates persistent notes so agents self-improve over time. Harvey reported a 6x jump in task completion rates; Wisedocs cut document review time by 50%. Available now in research preview alongside multi-agent orchestration in public beta.
OpenAI created the OpenAI Deployment Company — a TPG-led joint venture backed by $4B+ from 19 investment firms — to embed engineering teams inside enterprises for AI adoption. Simultaneously, OpenAI acquired Tomoro (~150 forward-deployed engineers) who've worked with Tesco, Virgin Atlantic, and Supercell on production AI workflows.
The DoD is deploying Anthropic's unreleased Mythos model under "Project Glasswing" to find and patch software vulnerabilities across government systems — even as Anthropic remains formally blacklisted as a Pentagon supplier. Pentagon CTO called it "a separate national security moment." Mythos can reportedly detect decades-old vulnerabilities in "minutes to seconds."
Former DeepMind RL lead David Silver's startup closed the largest seed round in European history at a $5.1B valuation. The "superlearner" aims to discover knowledge via pure reinforcement learning without human data — the same approach behind AlphaZero. Backed by Sequoia, Lightspeed, Nvidia, Google, and the UK Sovereign AI Fund.
Google quietly shut down Project Mariner on May 4, ending its 17-month experiment in autonomous web browsing via screenshot-based agents. The reason: visual processing at that scale was too compute-intensive and error-prone, and competitors like Claude Code made code-capable agents the new standard. Mariner's tech is being absorbed into Gemini Agent and Chrome's auto-browse features.
