Good morning. It's Wednesday, June 10, and we're covering xAI's unexpected partnership with Anthropic, the future of AI-powered warfare, and Anthropic's decision to lock down its most powerful models.

Plus: don't miss today's Real or AI challenge and check out the results of yesterday's poll on whether governments should own AI labs.

Anthropic's most capable AI arrives with built-in restrictions that intentionally route some risky requests to an older model. The company launches Claude Fable 5 for general use and Claude Mythos 5 for trusted cybersecurity partners through Project Glasswing. Mythos 5 pricing falls to $10 per million input tokens and $50 per million output tokens, while Stripe says Fable 5 cut months of engineering work to days.

China plans to spend about 2 trillion yuan ($295 billion) on a nationwide AI computing network, signaling that infrastructure may matter as much as frontier models. The government is drafting a five-year blueprint for interconnected data centers run largely by state telecom firms. The plan targets 80% domestic technology, favoring Huawei over NVIDIA and AMD, and relies on sovereign debt plus private financing.

OpenAI is pursuing a $1 trillion IPO while reinventing ChatGPT as an AI superapp, betting that agents will outgrow traditional chatbots. The company has confidentially filed for a US listing and is merging products like ChatGPT and Codex under a unified platform. OpenAI reports $2 billion in monthly revenue, nearly 1 billion users, and 2 million business customers as it targets profitability by 2030.

Genspark went from zero to $250M ARR in 12 months—and it's easy to see why. It's an all-in-one AI Workspace powered by the models you already love (Claude, Gemini, GPT, Nano Banana Pro, GPT Image, Veo, etc) that actually does the work for you.

One prompt builds your slides, sheets, docs, designs, websites, images, videos. And in 2026, all paid users get unlimited AI Chat and AI Image. Less prompting. More shipping.

Try Genspark free—sign up and get free credits today.

"Loop engineering" lets AI coding agents work autonomously until goals are met, replacing constant prompts with self-running workflows.

The following essay was written by Matthew Berman, sharing his firsthand experience with Anthropic's latest model.

I've been testing Fable (Mythos) for the past week and it feels unlike any other model I've used. It feels, and is priced, like a next-generation model. It also has some real quirks.

Workflow mode is the standout. I told it to do a "full code review" and watched it spin up hundreds of agents in parallel. It assigned an individual agent to basically every file in my application. It found bugs, edge cases, missing documentation, and UX improvements all over my application. And I've given this same prompt to other models that didn't find nearly as many issues. → Read the full article here.

A report from The Information details how xAI spent months trying to use Anthropic’s Claude models to improve its own coding AI, including reported distillation projects and workarounds after Anthropic cut off direct access. The report says xAI also faced internal setbacks, including leadership turnover, engineering cuts, and the accidental loss of weeks of training data, while struggling to match rivals in coding-focused AI.

As development challenges mounted, xAI leaned on outside partners and ultimately struck an unexpected agreement for Anthropic to lease a large share of its computing capacity for $1.25 billion a month, even as Elon Musk insists the arrangement is temporary. The story suggests xAI’s vast GPU infrastructure may be becoming as strategically valuable as its AI models, raising fresh questions about whether SpaceX’s AI ambitions will center on building frontier systems or monetizing the hardware behind them. → Read the full article here. (Paywall)

A June 2026 BBC report examines Foundation Robotics, a San Francisco startup building humanoid robots for military use, from logistics and reconnaissance to potentially operating weapons. The company says its Phantom robot is being tested by the US and Ukrainian militaries, with US trials limited to handling tasks while Ukrainian testing includes weaponization, and it aims to produce 40,000 units annually by the end of 2027.

CEO Sankaet Pathak argues humanoid robots could reduce risks to soldiers and civilians, but experts caution the technology remains far from battlefield-ready, citing major hurdles in mobility, dexterity, battery life, and decision-making in unpredictable environments. → Read the full article here.

Ted Lieu Rejects AI Framework: House Democratic leader Ted Lieu slammed a bipartisan AI plan, saying it lacks support and leaves major policy gaps.

OpenAI Funds AI Economy Research: OpenAI will fund outside studies on AI's economic impact using privacy-protected data and research tools.

Orbital Bets on Space AI: Ex-Spin CEO Euwyn Poon raised $5M for Orbital to build space data centers powered by Nvidia chips and Starship launches.

Lovable Hits $500M Run Rate: AI coding startup says users built 50M projects, with 1M added weekly as non-coders create business software.

JPMorgan Expands Autonomous AI Agents: JPMorgan plans AI agents that can work for hours independently, building on tools boosting private banking.

Was this passage written by a human or AI?

Up where the Hufftuft trees huff in the breeze, past the Niblets that nibble on nubbins of cheese, there wandered a Whoozit with one purple shoe, who hadn't the foggiest notion what to do. "I could sit," said the Whoozit, "and count up my toes.

Or I could go GOING to wherever-it-goes!" So he wiggled both ears and he twirled his lone shoe, and found that a-going's a grand thing to do. For the trick, little friend, when you're stuck in a stew, is to do the next small silly something that's you.

Here’s how you voted:

🟨🟨⬜️⬜️⬜️⬜️ A government owning pieces of AI labs (17%)
🟨🟨🟨⬜️⬜️⬜️ Private labs answering to no one (18%)
🟩🟩🟩🟩🟩🟩 Both — too much power either way (39%)
🟨🟨🟨🟨⬜️⬜️ Neither — let the market decide (27%)

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team