OpenAI's agent army hits 10M users — double in two weeks
Codex and ChatGPT Work now count 10 million combined active users, up from 8M mid-month and ~1M in February.
Industry
Products, deals and adoption. 217 stories, page 5 of 10.
ChatGPT Crawls Back to WhatsApp After EU Slaps Meta
OpenAI's bot returns to WhatsApp in Europe — free, no login — and lands on Kakao and Viber too.
Apple sues OpenAI, says it stole hardware secrets 'at every level'
Apple's lawsuit accuses OpenAI of poaching engineers and lifting confidential device designs as it pushes into consumer hardware.
Google's new AI image model is called Nano Banana 2 Lite
Google's cheapest image generator yet: ~4-second renders at $0.034 per 1,000 images, built for high-volume pipelines.
Micron bets $250B on U.S. chips to feed the AI memory boom
Micron raises its U.S. spend to $250B through 2035, chasing 40% domestic DRAM output as AI devours memory.
Apple Sues OpenAI, Says It Stole the Whole Hardware Playbook
Apple accuses OpenAI of trade secret theft 'at every level' to build its Jony Ive AI gadget.
Anthropic tried to bill a free-tier user $16.6M
A Korean dev on the free plan with zero API usage got a $16.6M invoice — up 10x from $1.67M a day earlier.
Apple sues OpenAI over 'rotten to the core' hardware theft
Apple accuses OpenAI of stealing hardware trade secrets to build its AI gadgets — a stunning turn on its 2024 partner.
Google AI Studio apps get free ai.studio URLs
Apps vibe-coded in Google AI Studio can now claim personalized links on the ai.studio domain — no DNS setup required.
ByteDance's Seedream 5.0 Pro searches the web before it draws
ByteDance's new image model reasons over prompts and web-searches for facts before generating — a shot at GPT-Image 2.
ChatGPT Work's agent now runs in picture-in-picture
OpenAI's upgraded Computer Use lets you watch the agent click, type, and browse in real time — and it moves faster.
Muse Spark 1.1: Meta's first paid AI, built for agents
Meta opens its agentic model to devs with pay-as-you-go pricing, a 1M-token context, and parallel sub-agents.
Chinese AI models now eat up to 46% of US token usage
US firms are quietly routing nearly half their AI traffic to cheaper Chinese open models, up from 11% a year ago.
Meta's testing AI glasses that record your whole day
Prototype snaps a photo every few seconds and records audio nonstop — with no LED to warn bystanders.
Microsoft ditches OpenAI & Anthropic inside Excel and Outlook
Microsoft is quietly swapping in its own MAI models for routine Copilot tasks to slash its AI bill.
Samsung posts biggest quarterly profit in tech history
AI memory demand drives a 19x profit jump — $58.4B, the largest quarter any tech company has ever reported.
Samsung's 2026 profit could top its last 40 years combined
AI memory demand is set to push Samsung's full-year profit past its entire 40-year earnings history.
Claude's Voice Mode UI Leaks a Model Picker — Again
New Claude Voice Mode screens surface a model selector, refreshed voice samples, and that signature Anthropic glow.
Anthropic wants its own chip — and Samsung might build it
Anthropic is in early talks with Samsung to make a custom AI chip on a 2nm process. The full-stack race is on.
China's 360 Digital Builds AI Rival to Claude Mythos for Vuln Hunting
360 Digital's new AI Vulnerability Discovery Agent enters the autonomous zero-day race, directly targeting Anthropic's Claude Mythos benchmark.
Google: 75% of New Code Now Written by AI
Sundar Pichai confirmed at Cloud Next 2026 that AI generates three-quarters of all new Google code — up from 25% just 18 months ago.
Google: AI Now Writes 75% of Its New Code
Sundar Pichai revealed at Cloud Next 2026 that three-quarters of Google's new code is AI-generated — up from 50% just months ago.
Xiaomi's MiMo-V2-Pro: 1T params, $1/M tokens, 1M context
Xiaomi drops a trillion-parameter frontier model on OpenRouter at 97% lower cost than Claude Opus — the fourth Chinese lab to ship in a month.
Anthropic's Claude Mythos breached on its launch day
An anonymous Discord group accessed Mythos — Anthropic's restricted AI cybersecurity model — hours after announcement, using stolen contractor credentials.