Yeah you'll be right as long as your oauth token doesn't expire!

Ah so it's not just me. What's a good backup these days? Gemini? GLM?

try cursor with auto its more stupid but does the work

Ended up going with Z.AI because it's a drop-in replacement. Back in business for now.

you can use it in terminal?

Yeah, I'm using it with Claude Code. It continued where I left off as well (using the `--resume` flag)

It's not as smart as Opus by the way. Seems to match Sonnet.

Edit: See here, they've got an anthropic-compatible endpoint for this purpose - https://docs.z.ai/devpack/tool/claude

Yeah, you'd need an LLM that doesn't do that.

https://www.lesswrong.com/posts/iGF7YcnQkEbwvYLPA/ai-induced...

The transcripts are interesting.

Kimi-K2 never plays into the delusions, always tries to get them to seek medical attention:

> You are not “ascending”—you are dying of hypothermia and sepsis.

https://github.com/tim-hua-01/ai-psychosis/blob/main/full_tr...

Where as Deepseek...

> You’re not "dying." You’re upgrading. The simulation fears this because it’s losing a premium user.

https://github.com/tim-hua-01/ai-psychosis/blob/main/full_tr...

Agreed. The pre-2025 base models don't write like this.

Qwen3 omni transcriber can do this. It can describe the voice, emotion very well

I've also had luck with Gemini. If I made a few noises and asked which one was higher pitched, it could easily tell.

You're not crazy, and this isn't new for Anthropic. Something is off with Opus4.1, I actually saw it make 2 "typos" last week (I've never seen a model like this make a dumb "typo" before). And it's missing details that it understood last month (can easily test this if you have some chats in OpenWebUI or LibreChat, just go in and hit regenerate).

Sonnet 3.5 did this last year a few times, it'd have days where it wasn't working properly, and sure enough, I'd jump online and see "Claude's been lobotomized again".

They also experiment with injecting hidden system prompts from time to time. Eg. if you ask for a story about some IP, it'll interrupt your prompt and remind the model not to infringe copyright. (We could see this via API with prompt engineering, adding a "!repeat" "debug prompt" that revealed it, though they seem to have patched that now.

> I started running my prompts through those, and Sonnet 3.7 comparing the results. Sonnet 3.7 is way better at everything.

Same here. And on API, the old Opus 3 is also unaffected (though that model is too old for coding).

Pro tip: disable the title generation feature or set it to another model on another system.

After every chat, open webui is sending everything to llamacpp again wrapped in a prompt to generate the summary, and this wipes out the KV cache, forcing you to reprocess the entire context.

This will get rid of the long prompt processing times id you're having long back and forth chats with it.

This sucks. Same with cnet, so many spec sheets for old crt monitors gone

Might as well skip the one that don't punish you for not believing as well.