[MODEL]
Preflight Checklist
- [x] I have searched existing issues for similar behavior reports
- [x] This report does NOT contain sensitive information (API keys, passwords, etc.)
Type of Behavior Issue
Claude modified files I didn't ask it to modify
What You Asked Claude to Do
Summary
I am reporting a serious issue with Claude Sonnet 5. I spent almost €200 in usage credits, but the model consistently produced incorrect code and unusable outputs for my project. The AI failed to follow structure, generated invalid syntax, and repeatedly produced wrong logic.
Environment
- Model: Claude Sonnet 5
- Platform: VSCode AI agent / Continue.dev (or specify your exact setup)
- OS: Windows 11
- Project type: C++ (Ran Online server/client source)
Expected Behavior
The model should generate correct, consistent code, follow the existing project structure, and avoid hallucinated logic.
Actual Behavior
- Wrong variable names
- Incorrect logic in multiple modules
- Invalid C++/c# syntax
- Code that does not compile
- Repeated mistakes even after corrections
- Extremely high token usage resulting in ~€200 cost with no usable output
Steps to Reproduce
- Load the project in VSCode
- Ask Claude Sonnet 5 to generate or refactor code
- Observe incorrect output and repeated hallucinations
Impact
This caused significant financial loss and wasted development time. The model’s behavior is far below expected reliability for paid usage.
Additional Context
I can provide sample incorrect code, logs, or a short video if needed.
What Claude Actually Did
Summary
I am reporting a serious issue with Claude Sonnet 5. I spent almost €200 in usage credits, but the model consistently produced incorrect code and unusable outputs for my project. The AI failed to follow structure, generated invalid syntax, and repeatedly produced wrong logic.
Environment
- Model: Claude Sonnet 5
- Platform: VSCode AI agent / Continue.dev (or specify your exact setup)
- OS: Windows 11
- Project type: C++ (Ran Online server/client source)
Expected Behavior
The model should generate correct, consistent code, follow the existing project structure, and avoid hallucinated logic.
Actual Behavior
- Wrong variable names
- Incorrect logic in multiple modules
- Invalid C++/c# syntax
- Code that does not compile
- Repeated mistakes even after corrections
- Extremely high token usage resulting in ~€200 cost with no usable output
Steps to Reproduce
- Load the project in VSCode
- Ask Claude Sonnet 5 to generate or refactor code
- Observe incorrect output and repeated hallucinations
Impact
This caused significant financial loss and wasted development time. The model’s behavior is far below expected reliability for paid usage.
Additional Context
I can provide sample incorrect code, logs, or a short video if needed.
Expected Behavior
Summary
I am reporting a serious issue with Claude Sonnet 5. I spent almost €200 in usage credits, but the model consistently produced incorrect code and unusable outputs for my project. The AI failed to follow structure, generated invalid syntax, and repeatedly produced wrong logic.
Environment
- Model: Claude Sonnet 5
- Platform: VSCode AI agent / Continue.dev (or specify your exact setup)
- OS: Windows 11
- Project type: C++ (Ran Online server/client source)
Expected Behavior
The model should generate correct, consistent code, follow the existing project structure, and avoid hallucinated logic.
Actual Behavior
- Wrong variable names
- Incorrect logic in multiple modules
- Invalid C++/c# syntax
- Code that does not compile
- Repeated mistakes even after corrections
- Extremely high token usage resulting in ~€200 cost with no usable output
Steps to Reproduce
- Load the project in VSCode
- Ask Claude Sonnet 5 to generate or refactor code
- Observe incorrect output and repeated hallucinations
Impact
This caused significant financial loss and wasted development time. The model’s behavior is far below expected reliability for paid usage.
Additional Context
I can provide sample incorrect code, logs, or a short video if needed.
Files Affected
Permission Mode
Accept Edits was ON (auto-accepting changes)
Can You Reproduce This?
Yes, every time with the same prompt
Steps to Reproduce
_No response_
Claude Model
Sonnet
Relevant Conversation
Impact
Critical - Data loss or corrupted project
Claude Code Version
claude sonnet 5
Platform
Anthropic API
Additional Context
Summary
I am reporting a serious issue with Claude Sonnet 5. I spent almost €200 in usage credits, but the model consistently produced incorrect code and unusable outputs for my project. The AI failed to follow structure, generated invalid syntax, and repeatedly produced wrong logic.
Environment
- Model: Claude Sonnet 5
- Platform: VSCode AI agent / Continue.dev (or specify your exact setup)
- OS: Windows 11
- Project type: C++ (Ran Online server/client source)
Expected Behavior
The model should generate correct, consistent code, follow the existing project structure, and avoid hallucinated logic.
Actual Behavior
- Wrong variable names
- Incorrect logic in multiple modules
- Invalid C++/c# syntax
- Code that does not compile
- Repeated mistakes even after corrections
- Extremely high token usage resulting in ~€200 cost with no usable output
Steps to Reproduce
- Load the project in VSCode
- Ask Claude Sonnet 5 to generate or refactor code
- Observe incorrect output and repeated hallucinations
Impact
This caused significant financial loss and wasted development time. The model’s behavior is far below expected reliability for paid usage.
Additional Context
I can provide sample incorrect code, logs, or a short video if needed.