Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
singhrac
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
14 ms
·
151.
▲
by
singhrac
3y ago
Also take a look at Stalwart: https://github.com/stalwartlabs/mail-server I haven't used it, but have used some of the sub-crates for some email-parsing problems, and it worked great.
152.
▲
by
singhrac
3y ago
Very sad to see where this company has come since Hewlett & Packards “the HP way” years. Read The Big Score recently and was surprised to learn HP was once something different (my entire life they’ve made half-baked products that I coul
153.
▲
by
singhrac
3y ago
You’re joking, but I know some projects moving to Biomejs, because Prettier is now outdated (I don’t really write JavaScript, so I have no idea).
154.
▲
by
singhrac
3y ago
pg explicitly tries to write very simply, which I've definitely appreciated and tried to learn from. However, I think Patrick's unambiguous style is also interesting in its own way, and he is writing very specifically about how (h
155.
▲
by
singhrac
3y ago
What were the features that you needed from Obsidian?
156.
▲
by
singhrac
3y ago
Do you think you’d make a plugin interface to get books? For example from Standard Ebooks, via some sort of API? My guess is that they’d be ok with that usage if you ask and you can solve some cold start problems that way. And, of course, g
157.
▲
by
singhrac
3y ago
I’m not dodging the issue. You asked if finance played a useful role in society, and I gave concrete examples where it did. Those services are more profitable because they are more valuable, because they are now better at transferring risk.
158.
▲
by
singhrac
3y ago
shrug sure, but people don’t talk about it that way, at least around me. They use thermal to mean a power source where you burn something and get energy out.
159.
▲
by
singhrac
3y ago
Or any number of other ways to keep track of employee performance that don't rely on an inherently random outcome in an enterprise sales cycle? For example, responsiveness to inbound emails, outbound volume, demonstrated (lack of) know
160.
▲
by
singhrac
3y ago
Yes. You can argue that it shouldn’t pay so well, or that there are bad actors, but the financial industry DOES genuinely help society. Like all businesses, they provide services that people want. The easiest example is insurance (e.g. home
161.
▲
by
singhrac
3y ago
I think you’re being a bit pedantic, actually. I work in power systems in the US (though not an expert) and the term thermal being used to refer to coal, gas and nuclear, with the latter a bit flexible, is very common. For example, it’s ver
162.
▲
by
singhrac
3y ago
They're similarly expressive, and there are some areas where Polars (taking ideas from years of Pandas development) is ahead, for example with nullable dtypes. That being said, there are many, many times when I would be willing to rewr
163.
▲
by
singhrac
3y ago
I agree, I like being forced to make a decision on an activity sequentially. Prevents optimization, which is a common way to spin around doing nothing.
164.
▲
by
singhrac
3y ago
Texas exists at the crossroads of both the Midwestern Wind Belt (actually, it gets SE coastal wind too) and the Sun Belt. Also, it has large open areas that are sparsely inhabited, pretty much ideal for building large renewable power plants
165.
▲
by
singhrac
3y ago
Why is it too late? Plenty of founders are over 30. Even YC is happy to fund middle career founders rather than just college kids.
166.
▲
by
singhrac
3y ago
This is pretty disappointing to hear. I’m really surprised they can’t even get a clean build script for users, let alone integrate into the regular Pytorch releases.
167.
▲
by
singhrac
3y ago
Just to point out it does, kind of: https://github.com/intel/intel-extension-for-pytorch I've asked before if they'll merge it back into PyTorch main and include it in the CI, not sure if they've done th
168.
▲
by
singhrac
3y ago
Generating random numbers is a bit complicated! I wrote some of the samplers in Pytorch (probably replaced by now) and some of the underlying pseudo-random algorithms that work correctly in parallel are not exactly easy... running the same
169.
▲
by
singhrac
3y ago
Is $0.2-0.4/kWh a good estimate for price paid in a data center? That’s pretty expensive for energy, and I think vPPA prices at big data centers are much lower (I think 0.1 is a decent upper bound in the US, though I could see EU being
170.
▲
by
singhrac
3y ago
I definitely agree. There’s no reason in my opinion not to have an extension library that does good visualization via a force-directed graph or similar. The existing visualization methods are pretty barebones.
171.
▲
by
singhrac
3y ago
Ironically the Rio Grande Valley is also a major wind farm region (see https://eerscmap.usgs.gov/uswtdb/viewer/#6.94/27.473/-97.63 ) but I think the profits from oil and gas in the region are still higher
172.
▲
by
singhrac
3y ago
Lately I've been finding tldraw.com a better tool. I hope they don't mess up the tool with changes to the license.
173.
▲
by
singhrac
3y ago
I don’t agree with all of his points either but I think what he meant by that was a separate, semicolon’d statement that the interpreter is slow (not that typing is slowing it down). I think he’s saying that types are useful for performance
174.
▲
by
singhrac
3y ago
I get a lot of inbound in this form and after the first disaster I'm much more aware of what a nontechnical founder needs to be doing. I think part of the problem is that they imagine themselves as having CEO responsibilities at the st
175.
▲
by
singhrac
3y ago
One of the things I struggled with in Rust’s portable_simd is that I think fadd_fast and simd intrinsics don’t mix well. This makes writing dot products only ok, rather than really easy. Here’s (one) reference: https://internals.
176.
▲
by
singhrac
3y ago
A lot of CPU inference libraries (llama.cpp included) use as much SIMD as possible, sometimes by hand-writing loops. The one I hack on, llama.rs, uses portable_simd but specializes to your CPU at compile time. My experience has been that mo
177.
▲
by
singhrac
3y ago
I’m very bullish. My dad and I were talking about this the other day, but there are still so many quality-of-life improvements that are bottlenecked by not having enough developers or enough time. For example, deploying hardware to all su
178.
▲
by
singhrac
3y ago
You can use Gaudi2s at the new Intel Developer Cloud[0]. Not sure why don't offer it on AWS though, seems a bit odd since they have the DL1 instances for the first-gen Gaudis. [0]: https://developer.habana.ai/intel-deve
179.
▲
by
singhrac
3y ago
To elaborate: you can't really buy these except in specific configurations from Supermicro (usually 8x H100) or the like. So take whatever chip-specific cost you have in mind, and 8x it, and add on the cost of CPU/memory/stor
180.
▲
by
singhrac
3y ago
This looks really neat. I appreciate that you reference the previous work in this area (by josephg, Ink & Switch, Fugue etc.). I think the roadmap says that WASM is next as a target, and that makes sense for prioritization. Would you al
More ›