site banner

Friday Fun Thread for November 7, 2025

Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.

3
Jump in the discussion.

No email address required.

Anyone else finding the new Kimi to be kind of overrated, at least by the standards of 'wow closed source is fucked' sentiment I see on twitter? I did a couple of creative writing challenges and found it significantly inferior to Sonnet which is perfectly reasonable given the price differential. I gave Sonnet an example of one of Scott's 'house party in San Francisco' and tell it to write a similar one, without plagiarizing the ideas from the first (which AIs seem to struggle with given that if you fill up the context length and tell it to draw inspiration from without plagiarizing they struggle). Sonnet could do that, Kimi didn't. Sonnet knows what a text adventure is and lets the user fill in the actions for the character, Kimi will make up its own actions. It's logical abilities were pretty good though, somewhere around Grok 4 and Sonnet.

Is this another coding-maxxed model? I gave it a little drawing with css test and it wasn't as good as Sonnet and much worse than Opus 4.1. In short I guess I don't really believe in the benchmark figures and I certainly don't believe in 'Artificial Analysis' which just aggregates benchmarks together. Kimi is cost-efficient and pretty good but not highly performant I think.

Opinions on Kimi Thinking generally?

Interesting turns of phrase and very good for atmosphere (or at least the descriptions are novel for now) but it gets details wrong and steers all over the place.