@0xKNL__ It’ll just perhaps even reveal how many people actually just wake up and do things vs those who dont. Because now the tools would be available

@grugcapital they just didn’t have @/emperor ser, why must you hurt me this way

Imagine measuring impact factor of a paper on a scale of 5, sometimes you need trillions

@0xkrane Tragedy that this isn’t evident

RT @CantelopePeel: BUILD FOR PERFORMANCE, NO COMPROMISES. I know I have been off the radar up until recently, let's catch up. I've left F…

@toptickcrypto No real basis for gpt being heavily used for data gen, also oai doesnt publish the internal reasoning for its o1 series. Its true deepseek wont have the compute, but still has other ingenious ways (creating more verifiers for different domains etc)

@0xsmac @knowerofmarkets I’ve been stuck in the 4 figure hell for real. been thinking about this for quite a while, gotta tweet through

Not incorporating AI in your workflows/figuring out will become a strict disadvantage by the end of this year vs those who do.

@LacertaXG1 the paper is a delight to read, definitely worth taking a look. just for the aha moment section if nothing, its a very impressive feat. But afraid most people wont find the full use for it

really cool recent talk on agents (coding centric), recommend to anyone who wants a more deep learning/ai perspective, there's warranted and rightful skepticism about agents today, but i personally think the ceiling matters not the current baseline

@metatron_0x open world models arent very good, especially in changing metas, and even with the existence of bots people prefer playing against other people. Also increases overall skill level over time, post openai dota2, pros even adopted some of the things the model chose to do.

@__paleologo You might enjoy these talks if interested in similar work

can skip the last section if not mathematically inclined, pretty good read regardless. I've always thought being good at technical (or non-technical) communication and explaining things to others is very hard (and have tried to improve since).

missing degenspartan on the tl

@toptickcrypto I think the best indicator for this to some extent is tracking engineering job posts from top labs, they have unnerfed frontier models with them, at some point, they won’t exactly hire research engineers at the same pace after a while, except cracked ones.

@grugcapital The kind of degeneracy that led to 1990 japan would put even some or maybe even most crypto stuff to shame though

@izebel_eth There’s two things here, how much risk are you willing to take for 5-6%? Is it appealing to you today because its double digits? Aave/maker battle tested, but if it demand was dropping, these yields would be sub 8 in a few months