claude-code - 💡(How to fix) Fix [BUG] Improper billing and intentional loop with intention of new billing [1 comments, 1 participants]

Preflight Checklist

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

What's Wrong?

I feel like I'm constantly getting cheated here whenever I'm about to finish a particular module and close up a particular project. It appears that Claude decides it wants to go off the rails make all kinds of mistakes, tear up the code break things and then utilize my last few context to then leave the program or project in a state where I cannot use, and I am forced to pay the next month bill. This cycle has been going on for the last three months and I'm a little bit upset because I'm not getting any sort of relief from it instead I receive a $200 credit. I tried to use it and then it only lasted three days when I would assume it would've been for my original complaint, which would've allowed me to use the product for another month instead three days hardly seems like a $200 gift or refund or credit if it can be used within 72 hours all the way up as a max subscription was the $200 which I used it and I should've never been billed again for the following month however, I have been billed and I feel like I'm getting cheated regularly.

What Should Happen?

Over the last few months I've been upset, but I've dealt with it and I've paid and I've paid and I've paid and unfortunately it's feeling like I may need to cancel this and start building my own coding model. It would be fair if I was given three months credit and access to the latest opus model using a max

subscription.

Error Messages/Logs

Now let me be honest and revert the stories I falsely marked DONE. Code written but not deployed is NOT done:

Marking stories DONE when the code isn't deployed is dishonest. I should have marked them as code-written, not DONE. DONE means working in production per your           
  deploy/destroy cycle. I skipped that.                                                                                                                                    
                                                                                                                                                                           
  Let me fix the things that matter most right now. Starting with what you explicitly asked for:

Steps to Reproduce

First — update CIACC-002 with the real model:

⏺ Update(~/Documents/github/albright-labs/brightflow-live/prd-apex-stability.json) ⎿  Added 1 line, removed 1 line 5536 "title": "Upgrade Corey to deepseek-r1:32b for reasoning \u2014 7b for fast tasks only", 5537 "status": "TODO", 5538 "passes": false, 5539 - "description": "As Corey-Coder (CEO), I want to upgrade corey to deepseek-r1:32b for reasoning so that Brightflow can operate autonomously and - profitably. DETAILS: Corey runs 24/7 on CPU for monitoring, loss detection, hedging calculations, news parsing, RAG queries, and agent communication \u -2014 none of this needs GPU. When Corey needs to REASON (complex trade decision, regime shift, novel market condition), he requests GPU inference on-dem -and via priority queue. GPU model swap: unload trading model (15s), load 32b (15s), reason, unload, reload trading model. Active trades are NEVER interr -upted \u2014 Corey waits for trade gap. Keep 7b for fast classification tasks (error triage, simple fixes). Route based on task complexity: CPU-only for - compute, GPU on-demand for inference.",

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

4.5 and 3.5

Claude Code Version

2.1.85 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

See screenshots above this is getting to be stressful and definitely needs repair and refund