I’ve been relying on GitHub Copilot for a few months, and I’m becoming quite annoyed with how its capabilities seem to have fallen. A couple of months ago, I was successfully using it to craft intricate user interfaces and tackle difficult frontend issues. The suggestions were precise and genuinely helpful for advanced coding tasks.
Now, it appears that the AI has regressed significantly. I’ve been having trouble with the simplest style adjustments. Just yesterday, I wasted a lot of time trying to change a button color from purple to blue, and it couldn’t manage that at all. Such straightforward CSS tweaks used to be straightforward.
Has anyone observed this decline in performance? It feels like all models are underperforming simultaneously. I can’t help but wonder if an update has caused this issue or if I’m making some mistakes on my side.
github’s been pushing updates pretty aggressively lately and i think thats messing with the models. been coding for years and copilot definetly feels dumbed down recently. maybe they’re throttling the better models during peak hours? try switching between the different model options in settings - sometimes the legacy one works better than the newer ones.
totally feel u on this! copilot used to be super reliable, but now it feels like it’s not even trying. i got stuck on some simple stuff too. resetting the cache might help, but not sure. let’s hope they fix it soon!
I experienced something similar around three weeks ago where Copilot became almost unusable for basic tasks. What worked for me was completely signing out of GitHub Copilot and signing back in through VS Code settings. There seems to be some kind of authentication or session degradation that affects suggestion quality over time. After the fresh login, I noticed the suggestions returned to their previous quality level. Additionally, I discovered that having too many tabs open or working with very large files seems to impact its performance significantly. The decline you’re describing with CSS changes sounds exactly like what I encountered before doing the sign-out process.
I’ve noticed similar inconsistencies with Copilot lately, though my experience has been more intermittent than a complete decline. The quality seems to fluctuate dramatically between sessions. Sometimes it handles complex algorithm implementations flawlessly, then struggles with basic syntax the next day. I’ve found that being more explicit with context in comments before writing code helps somewhat. Also worth checking if your IDE extensions are up to date, as I discovered an outdated VS Code extension was interfering with suggestions. The model switching between different versions might also explain the erratic behavior you’re experiencing with simple tasks like CSS modifications.