Enterprises are stripping AI prompts down to broken, grammar-free language — dubbed "caveman" — to cut token costs. A senior OpenAI employee has contributed code to the project. It's a sign that the economics of AI at scale are still deeply unsolved.
The Department of Commerce reversed its export restrictions on Claude Fable 5 and Mythos 5, and Anthropic is restoring access. The brief ban rattled developers and exposed how quickly government policy can yank foundational AI tools out of production workflows.
Claude Sonnet 5 dropped with improved reliability for long-running agents — a known weak point in previous Sonnet models. Willison heads straight for the developer docs, where the real signal lives rather than the marketing announcement.
Also known as Gemini 3.1 Flash Lite Image, this model is engineered for velocity and scale in the Gemini API. Google is clearly targeting the high-throughput, latency-sensitive tier that competitors haven't fully locked down.
Henrico County, Virginia — one of America's densest data center hubs — is bracing for a 25% electricity cost increase and told school workers to conserve power. The infrastructure debt of the AI boom is starting to land on ordinary institutions.
eBay, Amazon, and Etsy are losing the fight against AI-generated seed scam listings featuring impossible, gorgeous blooms. It's a small but vivid example of AI-generated fakery flooding commerce platforms faster than moderation can respond.
Andrew Ng flags "loop engineering" as the breakout buzzphrase of the moment, tracing it to Boris Cherny (Claude Code's creator) and Peter Steinberger (OpenClaw's creator). The idea — designing feedback loops between AI agents rather than just prompting them — is quietly becoming the new unit of serious AI engineering work.
Ethan Mollick admits the Fable 5 ban delayed real work and forced him to defer projects. Coming from one of academia's most grounded AI researchers, this isn't hype — it's a data point that frontier models have become genuine infrastructure dependencies, not just productivity toys.
U.S. M2 money supply surged $247.8 billion in May alone, hitting a record $23.1 trillion — the largest monthly jump since May 2021. With Bitcoin sliding below $58K and ETFs shedding $4.5 billion in June, the macro liquidity story and the crypto story are clearly diverging in ways traders didn't price in.
The Fable 5 saga — banned by export controls, then restored within days — is the story that keeps giving. It confirmed something important: frontier AI models are now infrastructure. When governments can flip a switch and halt access to the most capable model on the market, every company building on top of that model inherits geopolitical risk they probably never modeled. The scramble to local AI wasn't paranoia. It was rational.
And then there's Henrico County telling teachers to close the blinds so data centers can keep the lights on. The energy bill for the AI boom isn't abstract anymore — it's in school budgets. As Bitcoin's worst ETF month on record collides with record M2 expansion, the macro backdrop for all of this is getting stranger, not clearer. The Nasdaq is up 2% today. Someone is very confident about something.