so if we could remove money from politics, and somehow direct it toward these problems efficiently, they could be solved

and we'd get better leaders!

I was going to point out that the scale of society-level problems greatly outstrips the typical campaign warchest

but looking a bit deeper, it appears that gross political spending is roughly the same magnitude as some homelessness solution estimates

it seems that any universal guidelines should be trained into the model, or may be eventually

I guess it's a good idea to be able to cleanly a/b test in order to tune

the real test is long term game-ability - looking forward to understanding how that can be addressed

pardon me if I'm mistaken, but isn't Zo more of a traditional walled garden?

sharing the good stuff, ty!

I suspect nearly all of the rails designed to help humans program are also helpful to LLMs

it will be interesting to find what kinds of things occupy each part of the Venn diagram

are there any systems that cycle between accumulating short-term stateful memory and periodically collapsing it back into the base model, analogous to waking/sleeping cycles in humans?

training arguably constitutes memory, it's just not as easy to update

I'd like to explore ways to collapse stateful memory back into the base model via some fine tuning, at least as a way to address memory bloat

it's odd that our education, economic, and social systems are able to rely on language for all of the heavy lifting, yet LLM naysayers are so eager to trivialize its power

wen "find friends by IP"

woo, just as we'll be finishing our migration to k8s!

until we break it?

vitriol powered pendulums swing higher every cycle

it seems like people tend to surround themselves with people that aren't particularly disagreeable

I'm very much a fan of building actionable signals, but there is some subtlety to human social perception

does it generally increase the quality of discourse when random people approach to express general disapproval?