Anyone else having serious quality issues with Claude lately?

I’m currently working on a website and two apps. One of them is a large app that requires a lot of development. Let me start with the app. The main thing I’ve noticed is that before this week, I could work comfortably on Opus 5.5 Medium (sometimes High) and barely made a dent in my usage. This week, I’ve hit the 5-hour limit every single day, and my weekly usage is already at 80% since the reset on Tuesday, and it was already at 70% before that. It’s insane. Before, I never even got close to 70% while using it way more. Yesterday I spent an entire session building out one part of my app, a tool that wasn’t working properly. After many hours, it finally worked well. Today I wanted to make a small tweak, and for some reason it broke that tool. I looked through it and couldn’t figure out why, so I asked Claude. It started investigating, claimed it hadn’t touched anything, and then spun up a new session to rebuild the whole thing from scratch. It spent 3 hours on it and left it even more broken. So I switched to Fable to investigate and roll it back to its previous state. It told me that “state” from yesterday didn’t exist, so it rebuilt it from scratch again. Another 2 hours. And once again, it doesn’t work properly. On top of that, today while I was making changes to some of my internal apps, it broke everything there too. Every single thing it implemented, it broke. It feels like an extremely cheap, outdated model. Same story with the website. I’m currently working on the branding and was exploring 3 art directions. The results felt like going back to one of those old, bad models. It didn’t follow any of my instructions, and they weren’t hard ones. Literally just: “Don’t use orange in the second version, use red. Use a different typeface for the third version” (and I gave it the exact one). It used neither red nor a different typeface. It burned credits generating really bad images that went in the opposite direction of what I asked for. Today was one of the worst days I’ve had with AI. I’m on the Max plan and I’m seriously thinking about cancelling for good. Not long ago I was using GPT with Astra, and I remember it took a big quality hit, which is why I came back to Claude. My first days with Claude and Opus 5.5 were great, but the drop in quality and the spike in usage, at least in my case, have been much worse than what I experienced with Astra when I left. Today it broke all my projects. And the tasks I was doing today were actually simple ones. Read more