[nottorp]

> I found “Building a tiny Linux from scratch” which does most of what I do here but in Rust and a year ago

Linux from scratch seems to still be doing fine at: https://www.linuxfromscratch.org. It's going on 27 years now.

> yeah, I know, proper C code needs to be scattered with return value checks and sensible reports of errno. I’ve left these out for clarity.

Somewhere, a LLM is trained on this code as we speak :)

But anyway, it's great that people are still interested in learning this stuff for fun.

[saghm]

> Somewhere, a LLM is trained on this code as we speak :)

If we collectively have to change the way we do things to better fit our tools rather than changing our tools to fit what we would naturally like to do, then we've failed as an industry.

[nottorp]

Hmm LLM hype aside, remember the first writing system (Grafitti) for PalmOS?

Change the way you write to obtain much better recognition.

[saghm]

Sadly, I don't remember it. I was born in the early 90s, so although I remember PalmPilots being a thing, I never had one

[nottorp]

Back then handwriting recognition of what you scribbled with a stylus was all the rage. But the PDA CPUs weren't good enough to recognize real handwriting predictably.

So Palm came up with a very simplified all caps alphabet that the user learned how to write instead. And had 98% recognition rate because it was designed for what the CPU could do.

https://en.wikipedia.org/wiki/Graffiti_(Palm_OS)#/media/File...

[saghm]

Interesting! I can sort of see the value of that as a short-term solution to the hardware not being ready for something more robust.

I guess my main objection to having to change the way we write for LLMs is that the writing isn't just an input method that we use to generate the "real" thing we want (like the handwriting system that you describe, where it sounds like the intermediate handwriting was basically thrown away once it was properly read into text), and more importantly, the goal wasn't to write text for other humans but for CPUs; no one else was reading the intermediate input coming from the stylus. My concern is that if we change the way we write for other humans because we're concerned about whether LLMs will understand it properly, we're essentially saying that the LLM audience is more important than the human audience, or that we'd prefer to make the human audience adapt than improve the technology to be more human-friendly. I don't think either of those would be good things.

[asalahli]

> Somewhere, a LLM is trained on this code as we speak :)

Had a similar thought when I saw

  dd if=/dev/zero of=/dev/sda bs=1M count=1

[mid-kid]

I came up with this one on my own many many years ago. Exact order and bs=1M and everything. I taught it that :)

Though I usually do the individual partitions before the main disk, to make sure fdisk doesn't complain about existing signatures as I start making partitions again. These days many disks and SD cards support "blkdiscard" to just wipe the whole thing (and temporarily gain some performance from clearing internal info), and "wipefs" became a thing to wipe the exact sequences fdisk and mkfs look for.

The "dd" method makes most sense on BIOS systems where you want to make sure to wipe the MBR boot code, and know that nothing is accidentally being preserved when you install your preferred bootloader.