[BUG] Regression de claude code?

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Hi Anthropic team and community,
I'm a professional developer who uses Claude Code in VS Code daily for various projects (business applications, home automation, trading scripts, Python development). I've been a paying user since the beginning and I wanted to share honest feedback.
What I'm experiencing
Opus 4.5 at launch (late November): Impressive quality jump. Complex code generation was reliable, context retention was excellent, and the model understood nuanced requirements.
Opus 4.5 now (December): Significant regression. I'm seeing:

Basic coding errors that weren't happening 2 weeks ago
Quality that feels closer to Sonnet than what Opus should deliver
The model taking shortcuts (commenting out problematic code instead of fixing it)
Loss of context mid-conversation more frequently

The problem with feedback
In Claude Code (VS Code), there is no thumbs down button to report bad outputs. This means you're missing critical feedback from your most engaged users - the ones coding with Claude every day.
My ask

Acknowledge if there are infrastructure issues affecting Opus 4.5 quality (like the TPU bugs confirmed in August-September)
Add a feedback mechanism to Claude Code
Be transparent about load balancing / throttling if it's happening

I'm not trying to complain - I genuinely want Claude to succeed and I'm willing to pay for quality. But right now, the premium I'm paying for Opus doesn't feel justified compared to what I experienced at launch.
Anyone else noticing the same regression?

What Should Happen?

Expected behavior for Opus 4.5 (premium tier model):

1. CONSISTENT QUALITY: Output quality should remain stable over time, matching

the performance level demonstrated at launch (late November 2025)

1. PROPER CODE FIXES: When asked to fix a bug, the model should analyze and

correct the code - NOT comment it out, delete it, or work around it

1. CONTEXT RETENTION: The model should maintain full context throughout a

coding session without "forgetting" previous instructions or code structure

1. OPUS-TIER PERFORMANCE: Clear quality differentiation from Sonnet. Users

paying for Opus should receive Opus-level reasoning and code generation

1. TRANSPARENCY: If infrastructure issues affect model performance, users

should be informed rather than silently receiving degraded outputs

Error Messages/Logs

N/A - This is not a crash or error message issue. 
This is about output quality degradation that cannot be captured in logs.

Steps to Reproduce

N/A - This is not a crash or error message issue.
This is about output quality degradation that cannot be captured in logs.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.0.62

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_