[BUG] Recent Significant Drop in Quality

Environment

• Platform (select one):
• [ ] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [x ] Other: Linux remote server
• Claude CLI version: output of 0.2.54 (Claude Code)
• Operating System: Ubuntu 22.04

Bug Description

On every project and in every case, Claude Code has recently become borderline unusable and is now second in line to Cursor agent which I had not used for a month prior.

Whereas maybe a few days ago and all before that, Claude Code was a genius, there is just simply no other way to understand what the hell has changed other than assuming perhaps a back end change to the intelligence of the underlying model that's used because surely other people are noticing this right?

Steps to Reproduce

1. Use Claude Code last week
2. Use Claude Code this week

Expected Behavior

Follow my directions and achieve moderate successes

Actual Behavior

I can't count the times that in the last two days my explicit instructions, absolutely no uncertainty, have been just absolutely ignored I mean I've never ever wanted to punch an AI in the face so many times and so hard and somebody is going to have to be held accountable in the post AI world because people are not going to live like this. We are not going to accept Rogues deciding that today's an easy day off in lazy AI world and that this thing is too ** lazy to read entire files in one go I mean I'm paying again i'm just going to say this again okay, 150 dollars a day, $150 a day, and this thing will not read 50 lines consecutively without being prompted to do so every ** time every time and even then it's like maybe it'll read the file I told it to maybe it won't, maybe it'll just keep doing what it was doing despite what I told it to do and whatever over engineering is going on has to stop this is just crazy.

For 150 ** dollars a day, it should literally be doing 1000 lines per request and it's so infuriating how there is no setting to make this ** thing absorb more context or remember that it needs to I mean do you realize that this thing doesn't even read Claude MD consistently? I have followed the read me and every piece of documentation to the T and I always have and yet somehow everybody is somehow able to get Claude dot MD to be the instruction manual here because I don't even think it's ever read it I mean every time I refer to it it says I have to actually read it for the first time

If you look at my Twitter you'll see all I talk about is what how great Claude Code was for months or a month or whatever, but I'm about to become its greatest enemy if my suspicions are true about the quality

Additional Context

<!-- Add any other context about the problem here, such as screenshots, logs, etc. -->