I’m seeing concerning behaviour with cloud Opus 4.8, skipping githooks because there’s quality gates there, changing line limits, always asking me to do /goal clear
#claude@claudeai@ClaudeDevs
Claude Opus 4.8 on Max, on fucking maximum effort, is the dumbest motherfucker I’ve ever seen.
It follows a fucking /goal, stops, detects there’s a bug, and then asks me how it should proceed?
Fix the bug or not? What the fuck do you think?
#claude@claudeai@ClaudeDevs
Im testing cloud opus 4.8 on a max plan on the new maximum effort level it’s just shit and slow !!!!! GPT5.5 it’s much faster and with the right harness outputs better code than opus 4.8 !!!
Codex continues to degrade in real time.
ChatGPT 5.5 is either been nerfed or under compute capacity constraints.
Responses are taking forever, it feels like Claude Code all over again.
That magic that 5.5 had, has gone.
Just me?
Oh my god it scored worse than Composer 2! Not even 2.5! And it cost 4x more to run!!!
This might be the worst major lab model drop of all time. Llama 4 tier. Insane.
They are now training local recursive language models inside python REPLs with reinforcement learning
Agents are divide-and-conquering long problems, and passing around context as python variables.
This is so cool