claude is not what it was saddened by noticiable decline

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 want to start by saying I LOVE Claude Code. It has genuinely changed my life. I'm
a non-technical user building a full e-commerce platform and mobile app almost
entirely through Claude Code — something I never thought I'd be capable of. This tool
has empowered me in ways I can't overstate. That's why I'm writing this — because I
care about it and want to see it stay great.

▎ Over the past 30-45 days, I've noticed a real decline and it's been bumming me out:

▎ Speed: Tasks that used to take seconds now take minutes. I watch the token counter
barely move while the clock keeps running — most of the wait is just dead time, not
actual thinking. Deep think mode and builder agents are especially affected.

▎ Quality: Claude used to follow processes start to finish without me having to
babysit. Now it skips steps, makes decisions on its own that it shouldn't (like
skipping setup to "save time"), and I'm correcting it multiple times for things that
used to just work. When it builds something wrong because it cut corners, I have to
go back and fix it — which takes way longer than doing it right the first time.

▎ Cost: I'm paying the same price for noticeably worse service. That's the part that
stings. I'm happy to pay for the tool I fell in love with — but that tool was faster,
sharper, and more reliable 30-45 days ago.

▎ This isn't a complaint from someone who wants to tear you down. This is from
someone who tells everyone they know about Claude and wants it to keep being the
amazing tool it was. Whatever changed recently — please look into it. The early-mid
March 2026 experience was the gold standard.

▎ With love from a loyal user ❤️

What Should Happen?

Claude Code should perform at the same speed and accuracy level it was at in
early-mid March 2026. Tasks should use tokens actively instead of sitting idle.
Workflows and processes should be followed completely without skipping steps or
requiring user correction.

Error Messages/Logs

Steps to Reproduce

use claude

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.104

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

get well