Overview
- Developers noticed Tuesday that Anthropic’s pricing page showed Claude Code excluded from the $20 Pro tier, and the company later said the change was an A/B test for about 2% of new prosumer signups before it reverted the public pages.
- Anthropic said existing Pro and Max subscribers were not affected, though some new users in the test could not access Claude Code, prompting online backlash and public jabs from OpenAI’s Sam Altman.
- Explaining the trial, Anthropic said subscribers now run multi-tool agents for hours, which drives far more compute and costs per user, so it is testing options like tighter usage limits or different plan inclusions.
- In a separate update, Anthropic published a post-mortem on recent Claude Code quality complaints, finding three product issues—a lower default ‘reasoning effort,’ a cache bug that wiped session “thinking” each turn, and a prompt change that reduced detail—that it has now fixed and followed with a usage-limit reset.
- The episode reflects a wider shift away from unlimited monthly AI plans, with GitHub pausing new Copilot signups and tightening caps as providers move toward metered or token-based billing to match heavy agent workloads.