Three months of user complaints met a consistent response: the degradation you're experiencing is an expected side effect of intentional product changes, and you can adjust your settings to compensate.
Boris Charny, who built Claude Code, initially explained the quality regression users had been reporting since February as the result of two deliberate decisions - hiding Claude's internal reasoning steps from the UI, and lowering the default reasoning effort level. Both were framed as intentional product choices, not problems. The message to frustrated developers was essentially: this is working as designed.
That position changed after a developer submitted detailed conversation logs showing specific failure modes. The transcripts documented instances where Claude Code's adaptive thinking feature - the system that decides how much reasoning effort to apply based on task complexity - was classifying file editing tasks as low-complexity when they clearly weren't. Charny reviewed the logs and updated his public stance: the flaw wasn't just about user-adjustable settings, it was in the adaptive thinking calibration itself.
What the Shift Actually Means
The difference between those two positions has direct consequences for users. "Adjust your effort setting" treats the problem as a configuration issue - the user's responsibility to manage. "There's a flaw in the adaptive thinking feature" means the default behavior had broken in a way that requires Anthropic to fix, not users to work around.
Charny's revised assessment was shared in a developer discussion, not in any official Anthropic communication - no status page update, no blog post, no in-product notice. Users who weren't following that specific thread still had no way to know their Claude Code had been underperforming for months, or that the initial "it's your settings" explanation had been revised.
There's no public timeline for when the adaptive thinking calibration will be corrected. Until then, the available workaround is manually overriding the default effort level - which is exactly the user-side fix Charny's first response pointed to, even though the cause has now been acknowledged as something Anthropic needs to address on their end.