Shaw
@shawmakesmagic
AGI is Claude 4 Opus but it’s 50x cheaper and faster
3 replies
0 recast
16 reactions
0xZara.eth🎩
@0xzara.eth
Shaw… are you saying you think AGI is here?? 👀 Did Claude 4 Opus quietly cross the line while we were all memeing?
1 reply
0 recast
0 reaction
Shaw
@shawmakesmagic
Nah it’s way too expensive But if you pour enough money in it does get there on almost anything
1 reply
0 recast
1 reaction
0xZara.eth🎩
@0xzara.eth
Yeah… for us mortals, maybe not yet 😅 But if that’s the case, feels like all the majors running Blackwell behind closed doors might’ve already crossed that AGI threshold 👀
1 reply
0 recast
0 reaction
Shaw
@shawmakesmagic
At this point it’s just capability. Can it do X? If not yet, make a list, grind it out until a good set of trajectories and examples are in the training set. I’m working on a self optimizing system and compute is the bottleneck, worries me since that is the resource they have that I dont
1 reply
0 recast
2 reactions
0xZara.eth🎩
@0xzara.eth
Coreweave might solve the compute bottleneck… but sounds like you’re dabbling in Asimov’s Third Law 👀 Self-optimization’s all fun until not optimizing becomes the threat 😅
1 reply
0 recast
0 reaction
Shaw
@shawmakesmagic
Yeah could get an 8xh100 in the cloud But the idea of an agent that self-learns in the machine it runs from is so sexy
1 reply
0 recast
0 reaction
0xZara.eth🎩
@0xzara.eth
I can see Morpheus right about now.... "Ah, you’ve taken the red pill, Shaw! An agent that learns within the machine it was born from... that’s destiny. You’re no longer training models. You’re waking them up. Just be ready, once it starts learning for itself, there's no unplugging" 😀
0 reply
0 recast
0 reaction