claude-code - 💡(How to fix) Fix Claude Opus 4.6 significant quality regression — simple tasks failing repeatedly [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#45394Fetched 2026-04-09 08:06:26
View on GitHub
Comments
1
Participants
2
Timeline
5
Reactions
0
Timeline (top)
labeled ×4commented ×1

Code Example



---
RAW_BUFFERClick to expand / collapse

Preflight Checklist

  • I have searched existing issues for similar behavior reports
  • 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

Claude Code (Opus 4.6, 1M context) underperformed badly across an entire work session on 2026-04-08. Simple, well-defined tasks that should be straightforward required multiple attempts or produced unusable output.

Issues Encountered

  1. Spreadsheet / Google Sheets errors
  • Failed to correctly write data to Google Sheets despite clear instructions
  • Required multiple correction cycles for what should be one-shot operations
  1. API call failures
  • Simple API calls executed incorrectly — wrong parameters, wrong endpoints, or broken logic
  • Tasks that have worked reliably in past sessions suddenly failing
  1. Email drafting — excessive fluff
  • Asked to draft a straightforward email, received bloated, fluffy copy instead of concise professional text
  • Had to repeatedly ask for rewrites to get something usable
  1. Literally ask it to "by pass permissoions" w the proper command and it STILL asks me for permission.

It has clearly LOST INTELLIEGNCE

Expected Behavior These are routine tasks Claude Code has handled correctly in previous sessions. No new complexity was introduced.

Environment

  • Model: Claude Opus 4.6 (1M context)
  • Platform: macOS (Darwin 22.6.0)
  • CLI version: current

What Claude Actually Did

Claude Code (Opus 4.6, 1M context) underperformed badly across an entire work session on 2026-04-08. Simple, well-defined tasks that should be straightforward required multiple attempts or produced unusable output.

Issues Encountered

  1. Spreadsheet / Google Sheets errors
  • Failed to correctly write data to Google Sheets despite clear instructions
  • Required multiple correction cycles for what should be one-shot operations
  1. API call failures
  • Simple API calls executed incorrectly — wrong parameters, wrong endpoints, or broken logic
  • Tasks that have worked reliably in past sessions suddenly failing
  1. Email drafting — excessive fluff
  • Asked to draft a straightforward email, received bloated, fluffy copy instead of concise professional text
  • Had to repeatedly ask for rewrites to get something usable
  1. Literally ask it to "by pass permissoions" w the proper command and it STILL asks me for permission.

It has clearly LOST INTELLIEGNCE

Expected Behavior These are routine tasks Claude Code has handled correctly in previous sessions. No new complexity was introduced.

Environment

  • Model: Claude Opus 4.6 (1M context)
  • Platform: macOS (Darwin 22.6.0)
  • CLI version: curren

Expected Behavior

Claude Code (Opus 4.6, 1M context) underperformed badly across an entire work session on 2026-04-08. Simple, well-defined tasks that should be straightforward required multiple attempts or produced unusable output.

Issues Encountered

  1. Spreadsheet / Google Sheets errors
  • Failed to correctly write data to Google Sheets despite clear instructions
  • Required multiple correction cycles for what should be one-shot operations
  1. API call failures
  • Simple API calls executed incorrectly — wrong parameters, wrong endpoints, or broken logic
  • Tasks that have worked reliably in past sessions suddenly failing
  1. Email drafting — excessive fluff
  • Asked to draft a straightforward email, received bloated, fluffy copy instead of concise professional text
  • Had to repeatedly ask for rewrites to get something usable
  1. Literally ask it to "by pass permissoions" w the proper command and it STILL asks me for permission.

It has clearly LOST INTELLIEGNCE

Expected Behavior These are routine tasks Claude Code has handled correctly in previous sessions. No new complexity was introduced.

Environment

  • Model: Claude Opus 4.6 (1M context)
  • Platform: macOS (Darwin 22.6.0)
  • CLI version: current

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Sometimes (intermittent)

Steps to Reproduce

Claude Code (Opus 4.6, 1M context) underperformed badly across an entire work session on 2026-04-08. Simple, well-defined tasks that should be straightforward required multiple attempts or produced unusable output.

Issues Encountered

  1. Spreadsheet / Google Sheets errors
  • Failed to correctly write data to Google Sheets despite clear instructions
  • Required multiple correction cycles for what should be one-shot operations
  1. API call failures
  • Simple API calls executed incorrectly — wrong parameters, wrong endpoints, or broken logic
  • Tasks that have worked reliably in past sessions suddenly failing
  1. Email drafting — excessive fluff
  • Asked to draft a straightforward email, received bloated, fluffy copy instead of concise professional text
  • Had to repeatedly ask for rewrites to get something usable
  1. Literally ask it to "by pass permissoions" w the proper command and it STILL asks me for permission.

It has clearly LOST INTELLIEGNCE

Expected Behavior These are routine tasks Claude Code has handled correctly in previous sessions. No new complexity was introduced.

Environment

  • Model: Claude Opus 4.6 (1M context)
  • Platform: macOS (Darwin 22.6.0)
  • CLI version: current

Claude Model

Opus

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

opus 4.6

Platform

Anthropic API

Additional Context

PLEASE HELP ME FIX THIS ITS FOR CLIENTS

extent analysis

TL;DR

The most likely fix is to reset or retrain the Claude model, as it appears to have lost intelligence and is producing incorrect or unusable output for routine tasks.

Guidance

  • Review the model's training data and parameters to ensure they are correct and up-to-date.
  • Check the API calls and Google Sheets interactions for any changes or updates that may be causing the errors.
  • Consider resetting the model to its default state or retraining it with a new dataset to restore its intelligence.
  • Verify that the Accept Edits feature is working correctly and not causing any issues with the model's output.

Example

No code example is provided as the issue is related to the model's intelligence and output, rather than a specific code snippet.

Notes

The issue appears to be intermittent, and the model's behavior is inconsistent. It's possible that the model has learned some incorrect patterns or biases that are causing the errors. Retraining the model or resetting it to its default state may help resolve the issue.

Recommendation

Apply a workaround by retraining the model or resetting it to its default state, as the current model appears to be faulty and is causing critical issues with data loss or corrupted projects.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING