blt

4 ms·

31.
by blt 4mo ago
All technology, no matter how undesirable it once felt, eventually becomes nostalgic for somebody.
32.
by blt 5mo ago
It's funny how they specifically use the phrase "output that follows real-world physics" to describe the marble rolling video. At the end of the zigzag track, the marble jumps up for no reason. In a couple of other places it
33.
by blt 5mo ago
Agree. California peaches are horrible. Southern peaches (Georgia, Carolinas) are indescribably better. I love California, but it's funny/sad the extent to which many Californians deny that other parts of the USA have us beat in s
34.
by blt 5mo ago
I'm not a Haskell programmer, but from my limited awareness: Wouldn't they want to encode the restriction that April 31 doesn't exist directly in the type system instead of using raw integers for the underlying struct?
35.
by blt 6mo ago
Routing games like Mini Metro/Motorways, Freeways, and Fly Corp are not quite puzzle games in the traditional sense, but I'd definitely encourage any puzzle fans to try them.
36.
by blt 6mo ago
I knew that adding sriracha and hoisin was considered somewhat gauche / insulting to the chef. But I was surprised to see no mention of Basil, since every pho shop in the USA provides it. Is that also nontraditional? I really like it..
37.
by blt 6mo ago
In the dev branch now! https://asahilinux.org/2026/02/progress-report-6-19/ I will attempt to use Asahi as my daily driver once this is officially released.
38.
by blt 6mo ago
The title creator is bad... as a total amateur who needs to edit a video every now and then, it hurt me.
39.
by blt 7mo ago
yes of course, in my field the arrays are large and we're doing nontrivial work per element.
40.
by blt 7mo ago
Surprised Python is only 21x slower than C for tree traversal stuff. In my experience that's one of the most painful places to use Python. But maybe that's because I use numpy automatically when simple arrays are involved, and the
41.
by blt 7mo ago
For me it was worse, the headphone port on my phones always eventually failed. Maybe these rich people replace phones too quickly to experience that. I use wired at home, where I'm not cycling the connection very much.
42.
by blt 8mo ago
What makes this different from linking to a random zip file somewhere?
43.
by blt 8mo ago
And their predictions about Go were wrong, because they thought the algorithm would forever be α-β pruning with a weak value heuristic
44.
by blt 10mo ago
An underrated aspect of Matlab is its call-by-value semantics. Function arguments are copied by default. Python+NumPy is call-by-reference; mutations to array arguments are visible to the caller. This creates a big class of bugs that is har
45.
by blt 11mo ago
Clarity in writing comes mostly from the logical structure of ideas presented. Writing can have grammar/style errors but still be clear. If the structure is bad after translation, then it was bad before translation too.
46.
by blt 1y ago
Yes. Generative AI is bad. Most of the general population already realizes it. Only the tech and computer science bubble remains optimistic.
47.
by blt 1y ago
This needs to be refined: f(x) is O(g(x)) if there exists some X >= 0 such that f(x)/g(x) is bounded for all x > X . Otherwise, we cannot say that 1 is O(x), for example.
48.
by blt 1y ago
The definition of big O notation is pure math - there is nothing specific to analysis of algorithms. For example: "the function x^2 is O(x^3)" is a valid sentence in big-O notation, and is true. Big O is commonly used in other pla
49.
by blt 1y ago
Yes, in most cases the reduction of supervised learning to optimal control is not interesting. We can also reduce supervised learning to reinforcement learning, but that doesn't mean we should use RL algorithms to do supervised learnin
50.
by blt 1y ago
Awaiting the rocket engine equivalent of the K20.
51.
by blt 1y ago
tangentially, does anyone know a good way to limit web searches to the "low-background" era that integrates with address bar, OS right-click menus, etc? I often add a pre-2022 filter on searches manually in reaction to LLM junk re
52.
by blt 1y ago
IDK, I think Apple creating its own laptop/desktop-class CPU was a pretty bold move with a huge payoff. It's less sexy than introducing an entirely new category of product, but it's not exactly risk-averse either.
53.
by blt 1y ago
I always liked the look of ACM journals/conferences more than other venues. Their template uses Libertine, so it is my choice too.
54.
by blt 1y ago
Sure, but in games sometimes improving the average case is less important than the worst case.
55.
by blt 1y ago
What is an acceleration force?
56.
by blt 1y ago
Understandable. However, I went down that rabbit hole once, and learned that even summing a 1D array with a GPU is nondeterministic (because it is broken down like merge sort and subject to the scheduler). I guess I assumed that practitione
57.
by blt 1y ago
I agree with this part of your response: If we were to require that distributed training generates the exact same sequence of weight updates as serial SGD on a single machine, then we would need a barrier like that. However, there is a
58.
by blt 1y ago
I can see how such a phenomenon could happen at the level of a single machine, but if we're using a whole data center full of GPU machines it should be possible to spread out those spikes evenly over time. Still weird that the article
59.
by blt 1y ago
What is causing demand bursts in AI workloads? I would have expected that AI training is almost the exact opposite. Load a minibatch, take a gradient step, repeat forever. But the article claims that "each step of the computation corre
60.
by blt 1y ago
yes

More ›