[BUG] Significant degradation in Claude model quality since February 2026

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?

I'm a Claude Pro subscriber using Claude Code for professional software development. I need to report a serious decline in model capabilities that has directly impacted my ability to complete projects.

Issues observed since approximately February 2026:

1. Tasks that took 30 seconds now take significantly longer or fail entirely - Basic frontend work (CSS, layout, simple sections) that previously worked on first attempt now requires multiple iterations
2. Increased bugs from AI-generated code - The model introduces new bugs while fixing others, creating a cycle of "lazy patching" that makes codebases worse
3. Self-audit scores of 4-5/10 - When asked to evaluate its own output, the model acknowledges the work is poor quality, yet produces it anyway
4. Layout-breaking changes - Code modifications frequently break existing styling and structure, even when explicitly told to limit scope
5. Inability to complete basic tasks - Simple website sections that should be straightforward are now unreliable

Impact:

• Paying for Claude Pro but unable to complete projects
• Work that was reliable 3 months ago is now unusable
• Loss of productivity and trust in the platform

Serious concern about business practices:

The pattern here is troubling: model quality degrades for paying customers, then a "new model" is announced at premium pricing that restores capabilities we already had. This looks like a deliberate scheme - downgrade what customers
are paying for, then charge more to get it back.

If this is the business model, it's not innovation - it's artificial scarcity. Customers paid for a working product. Degrading it to upsell a "fix" is predatory.

I'm not the only one experiencing this. The timing is too convenient to ignore. Anthropic needs to address whether this is intentional, and if not, explain what happened and make it right for affected subscribers.

What Should Happen?

1. Transparency - Public explanation of what changed in February 2026 and why performance dropped
2. Access to Opus 4.7 for existing Pro subscribers - If this model restores previous capabilities, it should be included in current subscriptions, not sold as an upgrade
3. Compensation - Refund or credit for months of degraded service that failed to deliver what we paid for
4. Commitment - A clear policy that paying customers will not experience intentional capability downgrades

Error Messages/Logs

Steps to Reproduce

run any simple front end tasks

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.19

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_