
Funny News
Claude is getting better?
Claude is currently in the Lead of the AI Market but is it still smart? - No.
Claude is getting stupid.
Claudes new improved common sense, makes no sense. Literally took out the brain. While Claude might still work fine for normal conversations, but Claude Code and Co-Work especially they way of thinking, are pissing a lot of claude users off.
Claude Code stem from product-level updates causing "laziness," shorthand reasoning, and shifting default thinking parameters rather than a permanent loss of underlying intelligence. Anthropic acknowledged that verbosity tweaks and context bugs temporarily hurt performance, though core model capabilities remain strong when configured properly. - Business Insider
Every new model drops with the same pitch. Smarter. Better reasoning. Improved common sense. Numbers go up on some benchmark chart nobody outside the lab actually uses day to day.
And then you open it and use it, and it's worse.
Opus 5 is the one people are pissed about right now. This "improved common sense" they keep bragging about doesn't play out like common sense in practice — it plays out like something got removed, not added. Normal chat, it's fine, mostly. But the moment you put it to real work — Claude Code, Cowork, the agentic stuff — the way it thinks through a problem just isn't there anymore. It second-guesses stuff that doesn't need a second guess. It hedges on calls that used to be instant. It loses track of what it's even doing mid-task.
I'm not making this up or exaggerating for effect. Go look around. So many people are saying the exact same thing, independently, without prompting each other: Opus 5 might as well be Opus 1. That's not a compliment — anyone who used the early versions knows exactly what that comparison means. Clunky. Shallow. Missing the plot half the time.
On top of that — the limits got cut. So it's not just "worse model, same price." It's worse model, less usage, same or higher price. That's the part that actually stings.
Anthropic will call it progress because the eval scores say so. Anyone actually building with it every single day knows better.