[Bug Report] Model performance degradation causing incorrect logic and answers

Bug Description
I am writing to tell you how unhappy I am with how Claude Code is working right now. I have used this tool since the research preview, and I have never seen the models perform this badly.
Today alone, I ran into two costly bugs. To be clear, these were not bugs in my codebase, they were bugs and failures in the Claude model itself. What upsets me the most is that this keeps happening. Lately, the model's logic and answers have been getting worse and worse.
Because I have been here since the beginning, I know how good Claude can be. But right now, the model's mistakes are costing me time and money. I use this tool for real work, and these recent problems are ruining the trust I have had in it.
Please find out why the models are breaking and fix them. I want to keep using Claude Code, but I cannot afford to deal with major model failures every day.

Environment Info

• Platform: win32
• Terminal: vscode
• Version: 2.1.246
• Feedback ID: 099b010e-142e-4660-afc8-83058c8e487c

Errors

[{"error":"TelemetrySafeError: VirtualMessageList: itemKeys/messages length desync (keys=215 messages=214 range=[165,215))\n    at aM (B:/~BUN/root/_98.js:23:62348)\n    at pg (B:/~BUN/root/_98.js:23:55694)\n    at Bl (B:/~BUN/root/_482.js:21:19871)\n    at Ud (B:/~BUN/root/_482.js:21:37949)\n    at Fv (B:/~BUN/root/_482.js:21:83864)\n    at IR (B:/~BUN/root/_482.js:21:82875)\n    at lm (B:/~BUN/root/_482.js:21:82706)\n    at Mv (B:/~BUN/root/_482.js:21:79256)\n    at Ae (B:/~BUN/root/_482.js:21:6248)\n    at Ee (B:/~BUN/root/_482.js:21:4783)","timestamp":"2026-08-26T14:04:02.759Z"}]