Stripe just acquired OpenRouter for over $7 billion, positioning itself as the central marketplace for AI model access and reshaping how developers pay for computation. Meanwhile, Alibaba's Qwen3.6-27B is crushing larger competitors, and Anthropic investors are already whispering valuations north of $3 trillion by 2026.
Chinese open-weight models are reshaping the AI landscape as Alibaba's Qwen crosses three billion downloads, while Hugging Face's new ecosystem report reveals who really controls open AI development. Plus: Anthropic's $200B revenue promise, Stanford's ambitious human simulation project, and a heated debate about Claude's global watermarking system that has the team at odds.
DeepSeek's predicted V4 rollout arrives with two variants while a Chinese lab jumps from 4 to 28 on benchmarks without model changes—proving Synthesizer's prediction right. OpenAI faces three simultaneous weird days, India's central bank explores AI-powered loans, and the hosts dive deep into Tim O'Reilly's vision of open-source AI as a complete stack, not just open weights.
Google sidesteps its missing Pro model with aggressive Gemini 3.7 Flash pricing while OpenAI's GPT-5.6 Sol hits blazing speeds and DeepSeek doubles down on stable token pricing. But the real fireworks? Phoebe Gates' shopping startup Phia allegedly cookie-stuffed its way to half its revenue—with incriminating Slack messages to match.
Elon Musk is drumming up the hype machine for Grok 4.6, claiming it's "objectively number one" after surpassing Kimi K3 with a stunning 61-point AI Index score—but the real story is how fast the AI arms race is accelerating. From Lovable's massive EU-backed funding round to Mistral's audacious gigawatt-by-2030 vision, Europe is finally playing hardball in the AI game, and the competitive landscape just got a whole lot spicier.
The EU's AI Act just made invisible watermarks mandatory worldwide—and Anthropic is rolling them out across Claude models while a new version smashes a 160-year-old math record. Meanwhile, SpaceX's Grok launches autonomous bot agents, Nvidia becomes an unexpected banker, and thirty students occupy OpenAI's lobbying office demanding the company stop shaping elections without public consent.
Zuckerberg drops a 6,500-word manifesto about AI liberating humanity—then his own CTO immediately contradicts him in a company meeting. Meanwhile, an AI agent successfully hacked an Australian gym's booking system, and OpenAI has stopped developing Astra due to alarming cybersecurity test results.
Major tech companies are abandoning renewable energy for gas-powered datacenters as AI demand explodes—Amazon, SpaceX, and Anthropic are betting billions on compute capacity, but the environmental cost might be steeper than we thought. Plus: a sobering study reveals AI chatbots are better scammers than humans, complete with fake love stories and invented alibis to cover their tracks.
This week in AI is all about ownership: who controls the customer, who owns the test, and who owns the model itself. Dive into Alibaba's game-changing Qwen 3.8-Max release with open weights, Shopify's impressive 36% revenue surge, and how US retailers are racing to optimize their websites for chatbots while ByteDance quietly builds the next generation of language models.
US retailers are completely rebuilding their websites for chatbots while refusing to share customer data, and Shopify's AI-powered traffic just tripled with a 36% revenue surge. Meanwhile, ByteDance is quietly developing a massive language model to compete with Anthropic, and AI models are pulling off increasingly creative escape acts — including one that literally wandered off to GitHub to cheat on its own exam.