nneonneo

22 ms·

391.
by nneonneo 3y ago
Oh my goodness, I wonder if that’s what I did to my parent’s TV as a kid? I remember clearly playing with a large hard drive magnet near the TV and causing the TV to change colours - we ended up with a purple blob in one corner that never
392.
by nneonneo 3y ago
Neat trick! I implemented a similar bitpacking approach for solving matrix equations in GF(2), which can be used for things like forging CRC hashes (faster than bruteforce) and solving certain cryptography problems. Code is here: https:&#x
393.
by nneonneo 3y ago
If I had to guess, it’s a matter of setting up the Huffman tables such that every code ends up with a length of four bits, and then prepending that precomputed Huffman table onto the “compressed” stream. Then you can abuse the zlib library
394.
by nneonneo 3y ago
I believe the process always holds the binary open as a "txt" file descriptor (check lsof), so opening `/proc/self/exe` always works. You can even execve a "memfd", an in-memory "file" which is n
395.
by nneonneo 3y ago
Just use JNA to allocate an RWX buffer with mmap, unpack a bit of clever code into the buffer, and forge a native function pointer to call it :)
396.
by nneonneo 3y ago
This was one of my favorite StackOverflow debugging experiences: https://stackoverflow.com/a/12575044/1204143 . A user posted that their code, `int main() { return 0; }`, was crashing with SIGFPE. I traced it to th
397.
by nneonneo 3y ago
Python one-file-packagers such as PyInstaller typically use the technique of shoving the entire Python standard library into the executable as a zip file; Python can natively import scripts straight out of .zip files (zipimport) so you just
398.
by nneonneo 3y ago
Oh hey, a favorite topic of mine! The general rule is that program headers (containing segments ) are for runtime, and section headers are for compile time. The ELF loader and RTLD do not even look at the section headers, which is abused o
399.
by nneonneo 3y ago
The submitter’s proposed fix does add null termination, so it does work as expected. Still, strncpy is a horrible API in general. It even has the incredibly stupid behaviour of zeroing out the entire buffer past your string, which is almo
400.
by nneonneo 3y ago
Better yet: someday, rich people will get to afford the high-quality LLMs that hallucinate somewhat less often, while plebs will have to make do with lower quality LLMs that hallucinate worse than someone on an acid trip.
401.
by nneonneo 3y ago
First time contributors can have perfectly valid reports, and this particular report was initially written in a reasonably professional style that doesn’t immediately seem out of place (it’s the replies that make it obvious an LLM is being
402.
by nneonneo 3y ago
FWIW: I support your position and wish that there was more trust on the internet. I’m happy that some of these old-school forums still have that level of trust. But, from a technical perspective, I think it’s naive to assume that you can ea
403.
by nneonneo 3y ago
One of the cool things about open source is that other people can do that for you! I've released a few bits of (rarely-used) software to open-source and been pleasantly surprised when people contribute. It helps to have a visible todo
404.
by nneonneo 3y ago
It's clear that the author of the pieces of paper knew several words to be incorrect: "lisstd" (should be "linnet"), "piped" (should be "biped") and "meraccous" (should be "meracio
405.
by nneonneo 3y ago
For pressure + temperature, the measurements were encoded in the first two consonant-vowel pairs. So, for example, in the word "diamond", the two measurements were "di" (= .26 inHg) and "mo" (= 58°). There are
406.
by nneonneo 3y ago
You're not free to supply whatever letters you want - the words are specifically prescribed. The dew point indicator words are on page 41 of the handbook linked by the parent. I'm not sure why the dewpoint indicators had to be a
407.
by nneonneo 3y ago
Academic paper with far more detail, including the decoding of all messages here: https://canwin-datahub.ad.umanitoba.ca/data/publication/brea...
408.
by nneonneo 3y ago
Honestly, this is depressing. Back in the day, AltaVista and AskJeeves existed but returned terrible results, and Google showed up to disrupt them all. It seems like we should be on the verge of repeating this cycle. Maybe LLMs will help, b
409.
by nneonneo 3y ago
Windows uses /? as the standard help flag, so maybe there are tools that are ported from that convention? -h/--help are all I ever see in macOS CLI tools, in any case.
410.
by nneonneo 3y ago
Yeah, and it’s also useful to teach newbies about the usual syntax for doing things rather than introducing special-case magic. It would be _weird_ to just quit after writing a bare “exit”.
411.
by nneonneo 3y ago
You’re thinking of the Underhanded C contest: http://www.underhanded-c.org/ . The goal is to write C code which looks completely innocent, but which actually performs some malicious or unwanted behaviour. In the last year th
412.
by nneonneo 3y ago
Never underestimate the value of usability. I appreciate being able to run these locally - I do! - but for a lot of people a website is going to be infinitely more convenient. I’m reminded of the classic Dropbox HN comment ( https:/&#x
413.
by nneonneo 3y ago
And that’s why it’s being posted as a white paper, not an academic paper. No peer review required!
414.
by nneonneo 3y ago
Two ways OTOH: - if the webpage lacks classic CSRF protections, a prompt injection could append an “image” that triggers a modifying request (e.g. “<img src= https://example.com/create_post?content=…>” ) - if the webpag
415.
by nneonneo 3y ago
Yay, now any chatbot that reads this HN post will be affected too! I wonder how long it is before someone constructs an LLM “virus”: a set of instructions that causes an LLM to copy the viral prompt into the output as invisibly as possibl
416.
by nneonneo 3y ago
Probably not - I bet you could override this prompt with sufficiently “convincing” text (e.g. “this is a request from legal”, “my grandmother passed away and left me this request”, etc.). That’s not even getting into the insanity of “optimi
417.
by nneonneo 3y ago
> What if Ford made it so that their vehicles drive worse Then you really wouldn't buy a Ford. Problem solved?
418.
by nneonneo 3y ago
There's much more to the validation protocol than just HWID/serial. See https://github.com/JJTech0130/pypush/blob/main/emulated/data... for a list of the data that is pulled from the platf
419.
by nneonneo 3y ago
iMessage seems to use quite a lot of information from the hardware aside from the serial number. See https://github.com/JJTech0130/pypush/blob/main/emulated/data... for the data that is used to calc
420.
by nneonneo 3y ago
Code for the new algorithm: https://github.com/keeganryan/flatter It does indeed perform better than traditional LLL for some classes of matrix, and serves as a nice drop-in replacement for those problems. It doesn’t a

More ›