Going back to 4.6. Opus 5 is such a stupid being, that I don't want to talk to it at all. Why should I even talk to it?? It does not follow anything I say, it does not read it's rules, it doesn't read the code in the amount needed, after one prompt it again forgets what I wanted - it's, sorry, shit.
Just use the 4.6 again:
Claude --model claude-opus-4-6[1m] when you start it and /model claude-opus-4-6[1m] for in session setting of the opus 4.6. What I learned is if one is creating dynamic workflow and let Claude write it, then execute it, keeps the agents spawned in rails much better than opus 4.6 during development alone.
I also press 1 ("bad") every time the survey pops up in Claude code. It is performing BAD!
https://twitter-thread.com/t/2084727476761083926
Going back to 4.6. Opus 5 is such a stupid being, that I don't want to talk to it at all. Why should I even talk to it?? It does not follow anything I say, it does not read it's rules, it doesn't read the code in the amount needed, after one prompt it again forgets what I wanted - it's, sorry, shit.
Just use the 4.6 again:
Claude --model claude-opus-4-6[1m] when you start it and /model claude-opus-4-6[1m] for in session setting of the opus 4.6. What I learned is if one is creating dynamic workflow and let Claude write it, then execute it, keeps the agents spawned in rails much better than opus 4.6 during development alone.
I also press 1 ("bad") every time the survey pops up in Claude code. It is performing BAD!