Hacker Newsnew | past | comments | ask | show | jobs | submit | qihqi's commentslogin

prepositions p(a) -> q(a) can be thought as super set relationships. Let P = {a | p(a) holds} and Q = {a | q(a) holds) then p(a) -> q(a) and the statement P is subset of Q is the same.


Probably should make this into a backend of torch.compile


Yes, it would be a lot of fun if MPK can enable torch.compile to generate megakernels. Torch-generated kernels are currently too slow for latency-sensitive workloads.


For pytorch users: checkout the sister project: https://github.com/google/jetstream-pytorch/blob/main/benchm...


This artificial womb thing it talked about might be the solution to truly eliminate gender inequality. Oh Brave New World!


One issue I found with computer based setup when working is that... there are 2 of them.

When I was in college I had folders of notes and it's read and updated regularly. Once I got issued a work laptop; the personal one sometimes don't get turned on in days... and rapidly became outdated.


> there are 2 of them

Do you have a phone (doesn't have to be one you're using)? At least on android, you can install syncthing and use it as a middleman. I use emacs org mode for many of my notes and this system works well. My computers are NEVER online at the same time but the phone is always on so it will always be up-to-date with the latest files. It also serves as another backup location. Works wonders.


I use syncthing to keep notes from Boostnote (as well as other things) in sync across multiple computers. I imagine the same would work well with emacs org mode.


Population: One?


Since no one mentioned so far: The LOD law suit made FB unable to get any approved green card app since last few months of Trump. This is the top reason why Indians & Chinese are not taking FB roles. It has been recently resolved. If hiring gets better for them I'd attribute it to green cards than the rename.


Premature optimization is the root of all evil.


   (1) training ML systems on public data is fair use 

This one is tricky considering that kNN is also a ML system.


kNN needs to hold on to a complete copy of the dataset itself unlike a neural net where it's all mangled.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: