Anthropic introduces limits because of Chinese “hacker”

An incredible story about how one user hacked the artificial intelligence economy. And forced Anthropic to change the rules of the game for everyone.

A post appeared on a Chinese forum from someone who took 1st place on the Claude Count website. His expenses — $50,000 converted to API prices. And that’s a conservative estimate — the real sum could be 2 times higher due to data loading problems.

And here’s his formula for maximum consumption. Always use Claude Opus in ultrathink mode — this is the model’s deep thinking mode. Enable all MCP — model control protocols. Run multiple parallel copies of Claude Code on different projects simultaneously. This way token caching becomes minimal and costs skyrocket.

But the cleverest trick — the Background Commands function. It allows Claude Code to work 24 hours a day without stopping. The artificial intelligence literally doesn’t sleep, generating code around the clock.

The company realized — some clients use Claude Code practically without breaks, squeezing tens of thousands of dollars worth of value from a $200 subscription. And now Anthropic is introducing additional weekly limits for everyone. Because this smarty for $200 a month spent $50,000 worth of resources.

Most interesting — the author feels no remorse. On the contrary, he calls on others to follow his example. I quote — from August 28 limits are being introduced for Claude Code, only a week or 2 left, catch the last chance to go all out, let’s go.

This is how one user can disrupt the economy of an entire company worth tens of billions of dollars!