Developers and power users are accusing Anthropic of quietly degrading Claude’s performance, which Anthropic denies

A growing number of developers have taken to GitHub, X, and Reddit to claim that Claude Opus 4.6 and Claude Code have become less capable at sustained reasoning, more likely to abandon tasks midway through, and more prone to hallucinations than weeks earlier, with some framing the issue as “AI shrinkflation.” I've actually noticed it myself, thinking the other day, “WTF? Is this ChatGPT or something?” Anthropic has denied secretly degrading the model, with Claude Code lead Boris Cherny saying the most significant recent changes were disclosed product decisions — including shifting the default reasoning effort level to medium in March and hiding extended thinking output from the UI — not hidden quality reductions, and that users who want deeper reasoning can manually set effort to high. The controversy is playing out against a backdrop of real, confirmed changes to usage limits and cache behavior that Anthropic has acknowledged, which has made users more primed to suspect further undisclosed changes even as the company insists none have occurred.