Rem golden statue duping tech is actually unhinged
Helpers duplicate box souls + buffs and give you Alchemist-tier item timings on a 4. You become the one who knocks.
I don't think this is intended
The narrative that bigger models = better AI is dead. The next decade is about efficient algorithms on distributed compute. We've been investing in this thesis for 2 years. Most people are just catching up.
> prediction will be recasted as a hierarchical objective distributed across multiple levels of abstraction and temporal horizons along the depth of the model while being dynamically modulated to support optimally adaptive inference and learning
@yacinelearning Raw feeling is modern machine learning fails at distinguishing noise from signal so any technique relying on higher s:n is going to theoretically work better.
@iScienceLuvr My guess is you could distill the knowledge into a Lora of that size but it’d take a higher dimensional space to find a representation capable of being compressed to that degree
i dont understand how people dont outright recommend binary ninja over ghidra.
the ui, docs, and scriptability all seem to be 10x easier to learn and more longterm powerful
nothing liek the orgasmic release of getting your initial site load to be http3
https dns record: alpn="h3" port="443"
tcp has been removed from the stack
@davidyjeong@idro___ I feel like this is rarely discussed. Down to chat? Rebuilding an execution system to run a mm and arbs in the same process. mm runs every 1 micro and arbs every 2 sec so it’s been interesting