Context is fully broken

Context hasnt been working for days, it had been playing up before that but the last 2 to 3 days it is a liability to use Claude, especially with anything technical or complex or has code.
This is not context being used up then he is diminished, which is an ongoing problem, i mean after 1 context compression claude code should not be allowed to continue beyond that point, he is dangerous to the project, 100% everytime.
But for the last 2 to 3 days he is context damaged from the onset of fresh sessions, the very first response he is broken, you ask him to diagnose something non functional he merrily comes back and tells you its robust and 100 perfect.
These have been expensive weeks, you get a bit of the project to progress then spend twice as long repairing, if you can, the damage that claude has done, and you get trapped in a cycle of trying to get fresh sessions to get a fresh healthy claude to repair semi complex work, and the hole just gets deeper, and costlier, at our expense, its like Anthropic takes no care and consideration that they are charging us for broken services.But in the last 2 to 3 days, you start a new session thinking they must have either shut claude down, or fixed him, so you start new sessions, projects dont wait forever, and currently he has permanent ai dementia. And support is non existant.
Its a terrible situation and very very badly managed by Anthropic. And how they have designed it so claude will continue on after context compressions and overflows is bordering on, or is, unethical.