claude-code - 💡(How to fix) Fix Significant quality regression — paying subscriber requesting compensation [1 participants]

Subject: Significant quality regression in Claude — requesting compensation

I'm a paying subscriber and I've noticed a severe decline in Claude's capabilities over recent weeks/months. What used to be a highly capable AI assistant has become noticeably less intelligent, making basic logical errors and causing more problems than it solves.

Specific issues from today's session (2026-04-15, ~5 hours):

1. Claude killed a running ALVR Dashboard process "to read a config file" — reading does not require killing the process. This destroyed an active VR session that was working perfectly.

2. After breaking the working state, Claude made dozens of incremental changes trying to fix what it broke, each one making things worse. Instead of reverting to the known-working state, it kept experimenting.

3. Claude repeatedly asked me to test things instead of using available diagnostic tools autonomously. I had to constantly intervene and tell it to "use tools, be autonomous, find the errors yourself."

4. Claude suggested "is it late? should we stop for today?" multiple times despite being told not to — a pattern across multiple sessions.

5. Overall: tasks that should have taken 30 minutes took 5+ hours, with the end result being worse than where we started. The first fix worked within 8 minutes, then Claude's continued "improvements" broke everything.

This represents a clear regression from the quality I was paying for. I'm requesting either a refund for the affected billing period or account credit as compensation.