astroButter
@astroButter
An H100's worst nightmare בס״ד
So long as my mom still has to work I cannot stop grinding.
Me after a good nights sleep and 2 energy drinks
I’m pretty sure I’m top 5% life happiness right now regardless
My dawg with the banger of the year
I do not fear the man that has 10,000 vibe coded projects, but I do fear the man that has one project vibe coded 10,000 times ~Old Samurai saying
Average Meta Superintelligence NW: $100M Average Meta Superintelligence Deadlift: 20lbs Conclusion: minimize deadlift and you shall be rewarded.
chronicles of a wagie
i want to pee but i don't wanna leave my desk cuz that means my airpods will disconnect from my laptop
America was a superpower after WWII because we managed to attract every genius german scientist and researcher to invent incredible things in the USA and artificially increase our national IQ. This has led to an era of abundance in the US. We totally brain drained the world for…
Its really amazing how much effort goes into high level research as well as hardware design and firmware. Yet software to profile code gets almost no resources.
subhanallah for god did not know his creation would create something as incredible as pytorch chrome traces
Time to confess. I am on the epstein list. I was an MLE on the CUDA inference optimization team.
I think this is the tweet im most proud of.
If youre learning ML, forget it. Just get really good at writing fast gpu kernels. Become a gpu kernel monkey. Never import torch. Never import sklearn. Just make gpus go fast. You might ask yourself why you are making them go fast. Dont. Ignore the voices. Just make them go…
Social security is such an unbelievable scam I can’t believe it exists
Europe is ngmi because the damn European SWEs at my company which is a PRE REVENUE DEEP TECH STARTUP seem to think that it is okay to push a breaking change to main 5 mins before leaving the office and then go eat baguettes and drink wine like they haven’t just set all our…
If youre learning ML, forget it. Just get really good at writing fast gpu kernels. Become a gpu kernel monkey. Never import torch. Never import sklearn. Just make gpus go fast. You might ask yourself why you are making them go fast. Dont. Ignore the voices. Just make them go…