I guess that depends on what "really well" means

Such a great tool

This is fascinating, it begs the question: does the *structure* of a network provide any "self-knowledge" to the broader system?

I had a (maybe?) related thought recently, but with LLM networks, my response got too long so I made another thread about it

bsky.app/profile/cale...

More concretely, any LLM would, presumably, have some embedding vector for the word "grammar"; is there any mechanism by which the network, as it advances in response to a query *about* grammar, modulates its execution by connecting the word "grammar" to its own act of producing a grammatical?

Like, its whole *thing* is understanding language structure, in service of generating well-structured language. But can it ever articulate "meta-knowledge" derived from its network structure?

I guess I take solace that Zisner is still there, imo dividing up foundry and design is ultimately good for the product and seems like his doing

I'm genuinely surprised they didn't let him see his changes through to 2026, at least until the first process node under him comes out

BYU fans

I'm out of the loop, what changed? Did we lose our edge over the old NFL rules?

In that case it's in the hands of YCombinator now 🙏

I've been getting my ass kicked if you want some easy pickin' www.chess.com/member/caleb...

Hey Joris!

Does the PR build on your local machine? Or is it downloading a pre-built executable?

Heyo! I'm tryna see how many social networks I can connect to the same people on, here's to one more! 😅