Thx, that often works, but that's still inaccessible to me.

> Not a X. Not a Y. Not a Z. The thing he was most proud [...]"

This construction - how is it called? It is clear "AI speak" - at least I remember Claude talking like that (even if it is code).

If you think about it, it makes sense.

Say what you want about LLM-assisted software development, the chances are high that it will stay, meaning a non-trivial part of code will be written by an LLM.

So is it better to have

  - git commit (mostly only code)
  - magic or blackblock (meaning back and forth by developer with LLM, before commit)
  - git commit (mostly only code)
  ...
  (rinse and repeat)

or

  - git commit (prompt and/or code)
  - git commit (prompt and/or code)
  ...
  (rinse and repeat)

Obviously for that to work the LLM output has to be deterministic and a commit chain has to be pinpointed to a specific weight blob.

You can use ctrl+shift+t to open the recently closed tab again.

It's an interesting file browser, the default "blue" colors gives old-school vibes nowadays.

If you wonder how to quit (if started from a terminal):

It's `ESC 0`. Or "exit" like from a shell. Took me some time I have to admit (q, ctrl-c, ctrl-q, F10, ESC all did not work).

F10 always worked for me, but some terminals use F10 for some other functionality.

If that fails, you can click on "10 Quit" with the mouse (not ideal, but an immediate solution).

Ah I see, thx, it's a pretty default configured "gnome-terminal", which probably captures the F10.

Valid question i think in this context. I knew about GNU multiprecision library, but thought that couldnt be it, as it's "just" a highly optimized low level bit fiddling lib (at least thats my expectation without looking into the source), so it's strange why it could be damaging Hardware ...

TFA is on gmplib.org which kind of answers the question though.

It's okayish. Considering 64G to 128G are available for (nerd) high-end consumers you're just off with a factor 5 (if we can squeeze out a little bit more performance).

Thas is pretty astonishing in my opinion.

I think there is some kind of misunderstanding. Sure, if you get somehow structured, machine-generated PDFs parsing them might be feasible.

But what about the "scanned" document part? How do you handle that? Your PDF rendering engine probably just says: image at pos x,y with size height,width.

So as parent says you have to OCR/AI that photo anyway and it seems that's also a feasible approach for "real" pdfs.

Okay, this sounds like "because some part of the road is rough, why don't we just drive in the ditch along the road way all the way, we could drive a tank, that would solve it"?

My experience is that “text is actually images or paths” is closer to the 40% case than the 1% case.

So you could build an approach that works for the 60% case, is more complex to build, and produces inferior results, but then you still need to also build the ocr pipeline for the other 40%. And if you’re building the ocr pipeline anyway and it produces better results, why would you not use it 100%?

Do you think as an observer of Roko's basilisk ... should I share these prompt or not?