t55

7 ms·

31.
by t55 1y ago
Other debate topics can be found on the frontpage
32.
Show HN: Debate Uncle Bob – Is SQL Dead? (Voice RPG)(rehearsal.so)
6 points by t55 1y ago|1 comments
33.
OpenAI O3 and O4-Mini(openai.com)
1 points by t55 1y ago|0 comments
34.
by t55 1y ago
Congrats! obviously massive potential for saas automations what made you choose windows over other OSs?
35.
Memory in ChatGPT(twitter.com)
10 points by t55 1y ago|0 comments
36.
by t55 2y ago
Well, it depends. on some tasks, they surely are already super-intelligent.
37.
Superintelligence startup Reflection AI launches with $130M in funding(siliconangle.com)
38 points by t55 2y ago|26 comments
38.
by t55 2y ago
If they don't use some of the released kernels, then they are leaving money on the table
39.
by t55 2y ago
Never bet against a cracked team
40.
by t55 2y ago
in a parallel universe...
41.
by t55 2y ago
agreed, quite ambitious to cover this entire week in one post
42.
by t55 2y ago
The level of DS's openness has been really useful to actually understand what LLM folks are working on day in day out
43.
by t55 2y ago
Indeed. Modern LLM research is more about large-scale systems design than deriving gradients by hand. Math matters, but systems win.
44.
Intro to DeepSeek's open-source week and why it's a big deal(pyspur.dev)
24 points by t55 2y ago|13 comments
45.
by t55 2y ago
Anthropic doubling down on code makes sense, that has been their strong suit compared to all other models Curious how their Devin competitor will pan out given Devin's challenges
46.
by t55 2y ago
thank you!
47.
by t55 2y ago
hot take: i don't think you even need to understand much linear algebra/calculus to understand what a transformer does. like the math for that could probably be learned within a week of focused effort.
48.
by t55 2y ago
you're welcome!
49.
by t55 2y ago
Yes, it is great for key concepts but a bit outdated. Hence we added an LLM/FA section in the linked post!
50.
by t55 2y ago
I should have been more precise, sorry. Didn't want to imply they entirely ditched CUDA but basically circumvented it in a few areas like you said.
51.
by t55 2y ago
Agreed, not sure how much math is really needed.
52.
by t55 2y ago
lol. i guess this tutorial is about cutting out guido ;)
53.
by t55 2y ago
Hehe glad you did!
54.
by t55 2y ago
never heard of Hidet before; for when/what would I use it over CUDA/Triton/Pytorch?
55.
by t55 2y ago
ah gotcha. I think that with the new trend of RLing models, the move 37 may come up sooner than we think -- just provide the pretrained models some outcome-goal and the way it gets there may use low-level code without clean abstractions
56.
by t55 2y ago
interesting, you mean they are less obscure?
57.
by t55 2y ago
> Do you know when your docs will be a bit more comprehensive? Yes, we're actively working on this, and we should have some more pages by next week. If you have any questions, you can always shoot us an email: founders@pyspur.dev or
58.
by t55 2y ago
this looks really cool and i love rust. just a matter of time until everything runs on rust.
59.
by t55 2y ago
What do you mean?
60.
by t55 2y ago
Triton sits between CUDA and PyTorch and is built to work smoothly within the PyTorch ecosystem. In CUDA, on the other hand, you can directly manipulate warp-level primitives and fine-tune memory prefetching to reduce latency in eg. attenti

More ›