Google AI's frustrating limitations and why it fails
https://newatlas.com/ai-humanoids/exclusive-google-ai/

The AI responses suffer from a desire to please. Yes, decribing me as the Temu of AI is brutal but accurate.

I think the nugget there, and something I have read in more serious contexts, is that in general, applying "thinking" models is not useful for writing, summary tasks, but is useful for complex, multi-step problem-solving and research where you want sources verified.

QuoteHowever, when it comes to creative writing, human nuance, or following complex formatting commands, that extra processing often yields zero practical improvement. The system spends more computational power analyzing the data, but it still runs that data through the exact same restrictive output filters. The result is an illusion of depth that ultimately spits out the exact same generic, corner-cutting structure.

For churning out prose thoughts, it's a waste of tokens.