[BUG] 2.0.13 MASSIVE TOKEN USAGE DUE TO SYSTEM REMINDER
Status Fixed / completed
Maintainer reply None cached
Activity 5 comments · opened Oct 10, 2025 · closed Oct 15, 2025
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 had Claude editing a large MD file. It was reading just the bits it needed to edit. But EACH EDIT was eating 25 to 50% of the context WINDOW. Reverted to 2.0.10 and now each edit was taking 1% to 2%. Before I reverted I pointed out to Claude the context was being eating massively for no reason I could see and IT told me it was getting the COMPLETE FILE multiple times per edit VIA the SYSTEM REMINDER!!!!!!!!!!!!!!!!!!!!!!!!!!
What Should Happen?
NOT WASTE 25 to 50% of Context on system reminders?
Error Messages/Logs
Steps to Reproduce
Edit a large file with Claude in 2.0.13.
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.0.13
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
5 Comments
Found 2 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
I can confirm that. The context window is full after one or two prompts.
Dug into the jsonl logs some. Old logs ALSO displayed the "originalFile": contents when claude edits them. However before and after "cache_read_input_tokens" in those earlier logs only increased by a couple hunderd to a bit over 1000 tokens. Which was appropriate for the EDIT actually made in those cases. The 2.0.13 logs showed increases of 32000+ tokens between a single EDIT, showing the problem isn't in the system reminder messages themselves. It is in Anthropics TOKEN USAGE CALCULATION. It is/was INCLUDING the "oringinaFile": text as tokens ADDED to total token usage even though that file had been read as "originalFile" a dozen times in that session(and should not have increased token usage counts after the FIRST TIME IT WAS READ. So instead of going up 1k it was going up 30k, 60k or MORE depending how big the originalFile text was and how many edits were made to the file during the session.
Since you've ignored this issue. I've posted a new issue with EVEN MORE DETAILS and narrowed down the problem to a specific system reminder message so I am closing this ticket.
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.