Plus Google's secret Gemini chip, MCP's biggest update yet, and why the US Army is burning through its AI tokens
Sponsored by
Your next winning ad could be minutes away. AdCreative.ai helps founders and marketers create conversion-focused ads, visuals, and copy with AI. Try it free and launch smarter campaigns today.
Try it freeWelcome back to the world's #1 AI newsletter.
Today, we've got Anthropic's record-breaking $1.5 billion copyright settlement, China's massive open-weight models challenging US dominance, Google designing a chip that could make Gemini 10x more efficient, and the US Army literally running out of AI tokens.
It's been a wild week. Let's get into it.
On this day in 2025…
Google DeepMind unveiled Aeneas, the first AI model designed to help historians interpret, restore, and contextualize ancient Latin inscriptions. Built in partnership with the Universities of Oxford, Nottingham, and Warwick, Aeneas achieved 73% accuracy in filling textual gaps and could date inscriptions within 13 years of expert estimates. A reminder that AI isn't just about chatbots — it's rewriting our understanding of the past, too.
⚖️ Anthropic's $1.5B Copyright Settlement Gets Final Approval
A federal judge in San Francisco signed off on the largest known copyright settlement in US history. Anthropic will pay $1.5 billion to a class of authors who accused the company of using pirated books to train Claude. Around half a million writers are eligible for payments of at least $3,000 — but this doesn't resolve the broader AI training debate.
🇨🇳 China Drops Two Massive Open-Weight AI Models
Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-source model, while Alibaba previewed Qwen3.8 at 2.4 trillion parameters. Both companies are making their most capable models publicly available — a stark contrast to the closed approach of US labs. Demand was so high that Moonshot had to pause new subscriptions.
🔧 Google Designing a Custom Chip to Supercharge Gemini
Google is building a new server chip codenamed "Frozen v2" that would embed parts of Gemini's architecture directly into silicon. Engineers project it could serve 6-10x more tokens per unit of power than current TPUs. The catch? It's targeting 2028, and it only works if Google keeps Gemini's underlying architecture the same.
🔌 MCP Gets Its Biggest Usability Update Yet
The Model Context Protocol — the plumbing that lets AI models access your calendar, database, and tools — is getting a major upgrade. The new version switches to a "stateless" session approach, making it dramatically easier to run MCP servers at scale. This could finally unlock the wave of enterprise MCP integrations the industry's been waiting for.
🤖 Trigger.dev is an open-source platform for building durable AI agents and background jobs in TypeScript — with no timeouts, no infrastructure headaches, and serverless execution that scales. Think of it as the missing backend for your agentic AI workflows.
🧪 Supabase continues to be the open-source Firebase alternative everyone's building on. With Postgres, auth, storage, real-time subscriptions, and edge functions baked in, it's becoming the default backend for AI-powered apps — and it just keeps getting better.
Coming in Hot
Seedance 2.5 Create continuous 30-second native 4K AI videos
Seedance 2.5, Seedance 2.5 API, Seedance API, AI video API, AI video generator,
Your AI company brain that acts
AI voice and music production tools for creators
AI copilot for product managers
Full WordPress access for AI agents
AI video, image, and music generator for creators
AI-powered UI design for non-designers
AI-powered multi-exchange crypto trading terminal
Turn text into polished presentations with AI
Submit your AI tool to the #1 AI tools marketplace.
What it does: Seedance 2.5 is an AI video generator that turns text prompts and reference images into cinematic videos. It creates continuous 30-second shots with native 4K output, synchronized sound, and advanced scene consistency without requiring manual editing or timeline stitching.
Why It's Worth Your Time
Pricing
What it does: Kinovi is an AI video and image creation platform that lets creators generate visual content using multiple AI models from one workspace. It combines an easy-to-use Studio for no-code creation with an API for developers who want to integrate AI generation into their own products.
Why It's Worth Your Time
Pricing
🪖 The US Army Is Burning Through Its AI Tokens
Members of the Army received an email saying they were rapidly depleting their AI tokens and needed to limit use. Turns out even the military can't escape the "you've hit your usage limit" notification.
🤖 Gritt Exits Stealth with $34M for Construction Robots
Gritt is building AI-powered robots to automate the hardest tasks on construction sites, starting with solar plant assembly. With $34 million in funding, they're betting the future of construction is automated.
💰 Anthropic and Blackstone Bet Big on AI Implementation
"Ode with Anthropic" is the $1.5 billion AI implementation company that Anthropic launched with Blackstone, Goldman Sachs, and others. The bet? That the next trillion-dollar AI business isn't building models — it's helping enterprises actually deploy them.
That's it for this week. The copyright battles are heating up, open-weight models from China are shaking the industry, and Google is literally baking AI into silicon. If you found something useful in here, send it to someone who'd appreciate it. 🎉
Follow us on Threads, X, and Instagram for more AI updates.
Signing off,
— Somi AI
