winwang

26 ms·

391.
by winwang 3y ago
From my small personal perspective as someone who worked with ActivityPub a bit (side project), it's kind of an inefficient mess, no thanks to semantic web and JSON-LD (which, sure, it doesn't really need the full brunt of, iirc).
392.
by winwang 3y ago
Curious if this can be reconciled with the need for monetization. Do you know of propositions?
393.
by winwang 3y ago
Adding to my wikipedia reading list, thanks
394.
by winwang 3y ago
Over the past ~20+ years, the milk tea at NYC Chinese bakeries has gone from $1 to $1.10 to $1.25 to now $1.40.
395.
by winwang 3y ago
If it takes a few more decades, we could presumably find/implement better solutions.
396.
by winwang 3y ago
I agree. I see it all the time. Huge amounts of hours being wasted "working hard", and running away from answering the real questions which would double the (positive) output in half the time. That said, I think "laziness&quo
397.
by winwang 3y ago
I wonder if that's actually true, though. Especially since engineering is a creative field. For example, what if I churned out 20% less code, but bugs per line decreased by 50%? Or if we're architecting a system and the +50% weeke
398.
by winwang 3y ago
Sure, then I won't "press". I'll just release a product and see if the market wants it.
399.
by winwang 3y ago
considering that formal languages are part of math, there's little reason to believe natural language isn't. ft. LLMs.
400.
by winwang 3y ago
I agree with your sentiment, but want to point out the irony of choosing ignorance after being insulted as ignorant.
401.
by winwang 3y ago
I find it funny, even creative, that you managed to slip in something against tailwind in your comment.
402.
by winwang 3y ago
That's an interesting take. Crackpots and overhyped layman being a dual of anti-science. What would we call that? Well, I guess we'd normally talk about that stuff as science fiction.
403.
by winwang 3y ago
What I care about more than "fell for it" is the general lack of patience and skepticism from us, the audience and the commenters. As commenters, we are part of the media (i.e. social media). Usually, anything claiming itself a
404.
by winwang 3y ago
For reference, DDR4-3200 in quad channel is ~100 GB/s while a 3090's VRAM is 960 GB/s. Of course, most consumers only have dual channel. M1 Pro is 200 and M1 Max is 400. Which is slow for GPU memory, but incredible for main m
405.
by winwang 3y ago
Don't you also know the type of the object when dealing with pointers and references (assuming no casts)? Though arenas are useful even outside of "hacks". In the end, a pointer (or a reference, even) is "just" a us
406.
by winwang 3y ago
I think they meant that you'd expect simpler/more obvious ideas to be implemented first.
407.
by winwang 3y ago
I find that people place too much emphasis on "hard" work and not the "right" work. Personally, I know too many people who have worked their butts off... on the wrong problems. It's quite simple to continue sweating
408.
by winwang 3y ago
It's even worse in the case of hedge funds (well, arbitrage in general), because any given strategy has a maximum amount of profit it could make before the market loses that specific inefficiency. Or at least, that's the theory. I
409.
by winwang 3y ago
(Just an anecdote about my personal projects) I often switch between different ideas/features/projects in the same repo, and I like rebasing between different chains of thought. For me, the messages are more like a memorable phras
410.
by winwang 3y ago
(+1) I'm a newb to Metal myself, and I wanted to use Swift as the driving language (which was a main selling point). Unfortunately, almost all the material is in Objective C.
411.
by winwang 3y ago
If people like GPU programming, I wrote a blog post this week about GPU-accelerated hashmaps, semi-provocatively titled "Can we 10x Rust hashmap throughput?". HN post here: https://news.ycombinator.com/item?id=37
412.
by winwang 3y ago
At the same time, is there any reason to trust them over empirical evidence ( if there is evidence)? Incentives are incentivizing.
413.
by winwang 3y ago
Not the OP, but I'd presume that it would be significantly less tempting to sell out. Everyone has a price, and when everything is going smoothly, that price goes up.
414.
by winwang 3y ago
I'm not sure I understand. Is it that you don't want legal regulation to limit them (and similarly, other companies), but rather it were the market to naturally do so? I'm not sure that's sound economics. Or perhaps, tha
415.
Can we 10x Rust hashmap throughput?(wiwa.substack.com)
11 points by winwang 3y ago|0 comments
416.
by winwang 3y ago
- Add dependent types
417.
by winwang 3y ago
I think the idea is that, if we (some of us) can figure out when something is SEO spam or rather, generally low quality, an LLM should be able to but faster and more quantitatively why.
418.
by winwang 3y ago
I have found that most open source projects like this tend to oversell and underdeliver.
419.
by winwang 3y ago
And it's closely related to the Gibbs free energy (available energy), which decreases with increasing entropy, all other things equal.
420.
by winwang 3y ago
I'm not sure it's as simple as a "greedy corporation" thing, because they've been "greedy" for a while. In a sense, they're even legally obligated to be "greedy". And the more we skirt the d

More ›