The Stealth Experiment That Backfired
Anthropic quietly tested stripping Claude Code from its $20/month Pro plan, only to get caught red handed by sharp eyed developers. The company updated its public pricing page to show the agentic coding tool was exclusive to the $100/month Max tier, but only 2% of new signups actually saw the change. That didn’t stop the internet from roasting them. Users on Reddit and X rightfully called out the bait and switch: why update public docs for a micro experiment? The answer is obvious: they were gauging whether they could get away with it. Anthropic’s head of growth, Amol Avasare, eventually fessed up on social media, calling it a “small test.” But the damage was done. The pricing page has since reverted, but the trust took a hit.
The Real Problem: Claude Is Eating Itself Alive
Avasare’s explanation is honest, even if the rollout was sloppy. When Max launched a year ago, Claude Code didn’t exist. Neither did Cowork or long running async agents. Now users are running multi agent workflows that guzzle tokens like a V8 engine. Pro subscribers getting maybe 30 minutes of code time before hitting brutal weekly caps? That’s not a feature, it’s a rationing system designed to keep the lights on. Anthropic is scrambling because engagement per subscriber has exploded, and their compute budget can’t keep up. But instead of being upfront about capacity constraints, they tried a stealth nerf. That’s not how you treat developers who have built daily workflows on your tool. If Anthropic wants to survive the ChatGPT exodus, they need to stop playing whack a mole with pricing and start communicating clearly. Expect more outages and more limits until they figure out their infrastructure. Or until someone cracks the code (pun intended) on efficient agents.
Source: Arstechnica
