Recent searches
Search options
Someone used DeepSeek to write an optimization for DeepSeek that approximately doubled its performance: https://simonwillison.net/2025/Jan/27/llamacpp-pr/
For all the goofiness that accreted around the subject, this is literally what was meant by the technological singularity.
@tek I gave the distilled version a try, it's a lot more obvious how it's making decisions
@tek like, it "thinks" out loud a lot more than llama or gemma
@JoYo Definitely! And I find that as much or more interesting than its answers.
@tek I found it gets hung up on keywords kinda like Gemma. gemma2 always thinks I'm suicidal when I'm trying to work through thread locks.
@JoYo LMAO. That's amazing.
@tek I think this is more comparable to a compiler compiling itself, and then that resultant compiler is able to compile things faster. It doesn’t have the same ability to recursively self-improve like humans can (who, of course, cannot even do so indefinitely)
@tek have you tried it? It's still a pull request.
@Zeugs I haven’t.
@tek It's not real if you haven't run and at least tried a view prompts. 
Sorry not sorry.
@tek but will this optimisation finally tell us the what the Ultimate Question actually is?
@smiggs How many Matrixes must a man awaken from?