Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
haolez
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
22 ms
·
361.
▲
by
haolez
3y ago
Another way of seeing it is that "40% of the code became imaginary", in the sense that you have to figure out what the macro is expanding to. I'm not against generics. I was just wondering myself in a boring afternoon :)
362.
▲
by
haolez
3y ago
Can it be deployed as a web app? Might reduce the friction even further for new players
363.
▲
by
haolez
3y ago
It was even faster than that. I remember it being almost instantaneous. And you could even play Doom!
364.
▲
by
haolez
3y ago
I think TypeScript is more useful for guiding the IDE than checking the types. And that might be enough value to adopt it. It's just my feeling. But I don't feel like TS is preventing me to shoot my foot. The bugs that I write, ev
365.
▲
by
haolez
3y ago
I've tried Dex as a daily driver. It's unusable. It's slow, doesn't have customization options nor sane defaults. Alt-Tab is extremely non reliable. I ditched it.
366.
▲
by
haolez
3y ago
Yeah, but Textract uses OCR/computer vision even in PDFs with embedded text data and it can extract tables incredibly well. I believe there isn't an open source equivalent. Maybe some advanced usage of tesseract?
367.
▲
by
haolez
3y ago
That's awesome. I'm relying a lot on Amazon Textract for my PDF parsing needs. Do you have any other insights on how to do a good job at that natively, i.e. without a cloud provider? Especially when dealing with tables.
368.
▲
by
haolez
3y ago
Spyware sucks, but ransomware and joining a botnet suck much more.
369.
▲
by
haolez
3y ago
Good point. That's one problem that might be well solved by something like BitTorrent, which hashes all the files as part of the protocol. Once a binary gets analyzed and trusted (the current binary, not the process or every new upgrad
370.
▲
by
haolez
3y ago
Maybe I'm a bit paranoid, but I find it harder to trust the binaries generated by such a project than the ones provided by the Big Techs (even though they are the biggest stalkers). I'm afraid that such projects, with much less ma
371.
▲
by
haolez
3y ago
I see. But the cloud is much more than just VMs. I don't know if I'd want to manage an equivalent to SQS on my own. Maybe I should try it out and see what happens :)
372.
▲
by
haolez
3y ago
What about data center colocation? When you simply rent the energy, cooling, etc, but the hardware is yours? Do you think it's a nice middle ground?
373.
▲
by
haolez
3y ago
OpenAI's APIs error out a lot. You get a lot of 502 and other similar errors. This would probably affect someone trying to reproduce the paper. I think that's what they meant.
374.
▲
by
haolez
3y ago
I've just realized that MUDs can become and absurdly incredible experience with the advent of LLMs and diffusion models. You can drawn the most absurd and complex situations. The NPCs can be truly intelligent. Now I want to retire and
375.
▲
by
haolez
3y ago
On a related note, what's superior about Elixir compared to Erlang if we disregard Phoenix?
376.
▲
by
haolez
3y ago
What makes this game great besides the nostalgia factor? Genuinely curious. I play things like NetHack, so a good gameplay mechanic can go a long way with me :)
377.
▲
by
haolez
3y ago
From the data source at [0]: > Data are currently available from September 1, 1981. Seems like a short time range to analyze, but I guess most weather data is recent-ish. [0] https://www.ncei.noaa.gov/products/optimu
378.
▲
by
haolez
3y ago
I was thinking more about a recorder positioned on earth or on orbit, but I forgot about the main stage of the ship that deploys the probe and should have visibility of the dark side.
379.
▲
by
haolez
3y ago
On the dark side of the moon?
380.
▲
by
haolez
3y ago
Can I use this locally with models that run on my CPU? Like llama.cpp
381.
▲
by
haolez
3y ago
Maybe I'm weird, but I can imagine this being 10x more fun with a completely uncensored and toxic LLM backend :)
382.
▲
by
haolez
3y ago
They would just change the acquition details to avoid changing the owner and giving the account credentials over instead.
383.
▲
by
haolez
3y ago
I've looked into this recently and it's really good. They make use of Python's type annotations to do some clever conversions to frontend code.
384.
▲
by
haolez
3y ago
It seems almost too good to be true. Well done! Could someone share some bad experiences when adopting Nim so I can weight that in? I'm seriously considering it.
385.
▲
by
haolez
3y ago
My point is that the complexity is still there, but hidden. There is good and bad code in systemd as there is good and bad sysv init scripts.
386.
▲
by
haolez
3y ago
The bag of spaghetti in systemd's case is not visible to you, but it's there in thousands and thousands of C code (and I use systemd).
387.
▲
by
haolez
3y ago
Hypothetical future tax on pix transactions.
388.
▲
by
haolez
3y ago
I'm talking about a version of CPMF, but on Pix.
389.
▲
by
haolez
3y ago
Until they add a tax on Pix transactions.
390.
▲
by
haolez
3y ago
Thanks a lot! It does indeed look good. Btw, what made you choose Common Lisp instead of Scheme (Guile, Racket, etc) or Clojure? What made it more business effective? Genuine curiosity :)
More ›