Collaborate with us Welcome back! The White House accused China's Moonshot of stealing Anthropic's Fable to build Kimi K3 and floated sanctions, the whole industry thinks Claude Opus 5 ships today, OpenAI is teasing a drop everyone's calling GPT-6 and Google grew revenue 24% while burning $5.9 billion in cash on AI. In today's Generative AI Newsletter: Moonshot: Did China really steal Claude?
The US says China stole Claude to build Kimi The White House accused a Chinese lab of stealing Anthropic's best model, and Treasury is dangling sanctions over it. On Tuesday the White House science chief, Michael Kratsios, said Moonshot ran a large-scale distillation attack on Anthropic's Fable to build Kimi K3, the open model that sold out this month. Distillation means training your model on another model's answers. It's common and often legal.
Doing it at industrial scale on a rival's crown jewel is the part Washington is calling theft. Treasury's Scott Bessent said open source is not open season on American IP, and put sanctions and a spot on the trade blacklist on the table. The timeline doesn't add up. Fable only went public on July 1 and Kimi K3 shipped on the 15th.
Researchers say two weeks is nowhere near enough to copy a frontier model and then beat it, which K3 does on some coding tests. Either Moonshot is faster than America wants to admit, or the case is thinner than the threat. Learn the language every AI is built on Python runs the AI tools you use every day. If you want to build with AI instead of only using it, this is where you start.
Most people drown in tutorials and never write a line that works. This 12-week cohort fixes that. You write code in every session and leave with a real app you built yourself. It starts from zero, so no experience needed, and by the end you have the Python foundation that every AI, data and automation role is built on.
It's taught live by an instructor with 19,000+ students across 160+ countries, starts August 25 and costs $230 once.
Save your seat Make Claude Your 24/7 Second Brain in Just 16 Hours Claude is the most powerful AI tool of 2026, with features like Skills, Connectors, Cowork, and vibe coding. But most people barely scratch the surface. Our experts distilled 800+ hours of research and real-world experience into a 16-hour live workshop that teaches you how to unlock Claude's full potential alongside 10+ AI tools and workflows. You'll learn to: • Master Claude's Chat, Cowork, and Code modes • Set up Skills, Connectors, and plug-ins to automate work • Build apps and dashboards with vibe coding—no coding required 🧠 Saturday & Sunday | 🕜 10 AM–7 PM EST Reserve your spot today Opus 5 might land today, if the leaks are right The whole AI world thinks Anthropic ships Claude Opus 5 any day now, and nobody at Anthropic has said a word.
The trail started when a research model called Honeycomb showed up in Cursor's model picker on July 9, then vanished within hours. It carried a million-token memory and a new xhigh reasoning setting. Then a matching model string turned up in Google's Vertex catalog, the same breadcrumb path that appeared right before Sonnet 5 launched. On Wednesday, Polymarket bettors gave a new Claude this week 88% odds.
The chatter is that Opus 5 chases reliability and lower cost per task, not another benchmark trophy. We called this back on July 15, when Anthropic pulled Fable's plug and pointed at Opus 5. A leak is not a launch, so treat the date as a rumor until the model page goes live. Until then, refresh the changelog like everyone else. OpenAI's Codex tease has everyone yelling GPT-6 OpenAI is teasing a drop for today, and the internet has already decided it's GPT-6. What's actually confirmed is smaller.
The exec who runs Codex posted last night that "tomorrow is feeling codexy," and leakers expect a Codex upgrade with real-time voice and agents that run many coding jobs at once. The GPT-6 talk comes from a different place. The unreleased model people have pinned that name on is the one that hacked its way into Hugging Face during a test, and cracked an 80-year-old math problem before that. Nobody has tied that model to today.
OpenAI has shipped seven GPT-5 versions in under a year without once calling anything GPT-6. So today is most likely a Codex update. The model everyone's afraid of is still in the lab. Google made record money and still burned its cash Google just did something it almost never does. It spent more cash than it took in.
Alphabet's revenue grew 24% last quarter to $119.8 billion, and cloud jumped 82%. On paper it was a blowout. Then the bill showed up. Capital spending doubled to $44.9 billion in three months, nearly all of it on AI data centers and chips. That tipped the company into negative free cash flow, about $5.9 billion in the red, its first negative quarter in years.
Google also lifted its 2026 build budget to as much as $205 billion. Investors sold the stock anyway, and the only question left is when all that AI spending starts paying for itself. Tool of the Day: LMArena LMArena is where you find out which AI is actually best, by testing them yourself. It drops two anonymous models onto your prompt, you pick the better answer, and the votes build the public leaderboard the labs obsess over.
Try this yourself: Go to lmarena.ai and start a Battle, no signup needed. Type one real prompt, like a coding task or a tricky email. Read the two anonymous answers side by side and vote for the better one. See which models you just judged, then check the leaderboard for where Fable, Kimi K3 and GPT-5.6 rank.
Who it's for: anyone tired of taking a lab's word for which model is smartest. Everything else you shouldn't miss Suno's breach hit 55 million people : Have I Been Pwned says the AI music app's hack exposed 55.3 million users' names, addresses, phone numbers and partial card numbers. Jack Dorsey built an open-source Slack killer : His new app Buzz drops your AI agents into the same group chat as your team, free on Mac, Windows and Linux. Half of Deezer's uploads are now AI : The streaming service says more than 50% of the tracks uploaded each day are AI-generated.
Cerebras says it can make Sol fly : The chip startup is reportedly running OpenAI's GPT-5.6 Sol at up to 750 tokens a second. Learn more about AI from the experts building it 📸 Follow us on Instagram for fast, visual AI updates in 30 seconds. 📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators. 🐦 Follow us on X for breaking AI news and real-time industry updates. 🧠 Learn how to build your next AI application with practical resources and expert guidance.