HN user

antonly

115 karma

PhD Student in Compilers at Edinburgh University. Low-level shenanigans enthusiast.

Posts6
Comments30
View on HN

I wonder how common of a problem this will be in the future. The experiment will fail due to improper setup, the human will at best glance over the logs and declare victory, and everyone just believes.

How do we know free tier models will continue to provide adequate services? They are pretty good right now, but only because all big players are happy to burn enormous amounts of money.

But it won't even do that. In the pursuit of extracting maximal capital, better models with "better" knowledge will be gated behind higher prices. If you pay more, you get more.

How does it help equalise, if the commodity is sold "at market value" by the richest, and all intellectual input is immediately fed back into the tool. Looks at best like a weapon to supress

Alright, here I go.

I don't think you are AI. I merely lament the fact that you found it appropriate to post a clearly LLM generated comment.

The problem with LLM generated comments for me is not their content, but rather their nature. I am not addressing the "actual discussion", as in my personal opinion, there is no "discussion" to be had. The post constitutes an automated response akin to an answering machine (be it much more sophisticated), and I generally do not find discussions with answering machines interesting at all.

That's a major sticking point, but I am yak-shaving so hard that I ended up in a complete re-build of the static site generator and am currently in the middle of that... RSS hopefully coming in a few months once I find time to finish that process.

I honestly don't trust it for conflict resolution. I keep data that is edited in multiple places, potentially simultaneously off of it and use git instead. I've had one minor incident with corruption in the last ~year, although I already can't remember the exact circumstances. Large files are no problem at all, I've even streamed movies off of it.

Thanks, but I prefer open stuff (as stated in the article), so I'd rather use something like ollama or similar provider on a consumer-grade graphics card that I can use for other purposes too.

Some feedback on your website: The AI-generated cover image looks really awful, nothing makes sense. And your website makes my laptop sound like it's trying to compile LLVM. Both is a strong turn-off for me at least.

I wrote this as a rebuttal to the sentiment that "the open web is dying" that is often out forth here. Just accepting that walled gardens won is imho throwing the towel too early, as the open web is built by people like you and me, and hence only dies when we stop believing in it.

I'm sorry, I don't follow. Can you point me to where exactly I am "nitpicking" and not addressing the "main issue"?

Your main point was not "free movement of labor undermining sovereignty and eventually quality of living".

You started off by implying that the problems we see come from a "Import refugees and give them free money" policy.

I then pointed out that I don't think refugees receiving free money is the main problem, as there is comparatively little money going to refugees.

You then point out that 16% of the UK were foreign born, implying that they were refugees, that they get "free money", and that this is the reason why the UK has infrastructure problems now.

I simply pointed out that "foreign born" != "refugee".

If we entertain your goalpost-shifting and argue about free movement of labour as the root cause for lessened sovereignty and quality of living, I would like to ask you to how Chinese students coming to your country to study, or eastern European truckers trucking around your goods exactly undermine the sovereignty of your country. I am honestly curious.

I cannot figure out how this prevents your country from maintaining her own infrastructure.

Btw, I would invite you to reflect a bit on how you came to the extremely reductionist viewpoint that all foreign born people are refugees.

Woah there, not all foreign born people are refugees. That's some serious misunderstanding you have there.

Also, the "foreign born" statistic is pretty moot, seeing as UK was part of the EU until shortly before 2021 with free movement of labour. Most other European countries have higher levels of "foreign born" people living there, e.g. Austria, Germany and Sweden with around 20%[^1]. If you scroll further down, you see that ~half of these people are born within the EU.

So yeah, check your biases.

[^1]: https://ec.europa.eu/eurostat/statistics-explained/index.php...

Btw, I am "foreign born" within the UK, doing my PhD here in computer science. Most of my colleagues aren't from within the UK either. This is not because of any "policy", there simply is not enough demand from UK students to fill these places. Good luck with your country once you convinced all of us to leave. Have fun drinking your water :)

This is a really shallow and unfounded take. Sad to see this here.

This particular issue is imho mostly related to a lack in investment in water infrastructure (reservoirs and pipes). I don't see how migrants factor into this equation (not to mention that the "free money" given to migrants is scarcely a drop in the bucket). Please spread your hatred elsewhere.

This is actually a really nice example of how security tools can fall flat:

There is this check [here](https://github.com/TheAuditorTool/Auditor/blob/2a3565ad38ece...), labelled "Time-of-check-time-of-use (TOCTOU) race condition pattern".

It reads:

`if.\b(exists?|has|contains|includes)\b.then.*\b(create|add|insert|write)\b`

This matches any line that contains `if` followed by `has` followed by `then` followed by `add`, for example. This is woefully insufficient for actually detecting TOCTOU, and even worse, will flag many many things as false positives.

Now the real problem is, that the author states that this will solve all your problems (literally), providing a completely false sense of security...

TheAuditor solves ALL of this. It's not a "nice to have" - it's the missing piece that makes AI development actually trustworthy.

I've built the tool that makes AI assistants production-ready. This isn't competing with SonarQube/SemGrep. This is creating an entirely new category: AI Development Verification Tools.

Wow, that's a lot of talk for a tool that does regex searches and some AST matching, supporting only python and js (these things are not mentioned in the main project README as far as I can tell?).

The actual implementation details are buried in an (LLM written?) document: https://github.com/TheAuditorTool/Auditor/blob/main/ARCHITEC...

My favourite part is the "Pipeline System", which outlines a "14-phase analysis pipeline", but does not number these stages.

It reads a bit like the author is hiding what the tool actually does, which is sad, because there might be some really neat ideas in there, but they are really hard to make out.

PhD Timeline 1 year ago

Yes, thanks! English isn't my first language, and apparently I missed that there was a distinction between the two.

PhD Timeline 1 year ago

Speaking as an academic from the UK, there has been a growing sentiment of weariness about spending time in the US. From hallway jokes about getting snatched at the border and spending two weeks in detention camps, to people re-evaluating their conference and career choices to avoid the US, this marks a major shift in the mindset of the current generation of PhD students and early-career academics that will probably have untold implications for long-term research and innovation horizons...

I think part of the point is to find that problematic software, and contribute patches to make it work (or whatever you do for closed source software). In my mind this is a board to legitimise early adopters' requests, so that they can say "hey, this doesn't run on this very real laptop from a brand that you may have heard of".

I think this is a crucial step on the way to getting an actually good experience on RISC-V based platforms.

A while back I wanted to render some of these myself, the results can be seen at https://renders.datenvorr.at/. I have a few 760 megapixel versions of these that you can easily print at 1x1 meter with full detail if anyone is interested. I have a couple of these on my wall...

I used the project to familiarize myself with the C programming language, so the code quality is pretty bad. A while back I rewrote most parts of it, but never merged the finished result back into master...

I built a RISC-V emulator in python focused on extensibility and debugability for this exact purpose. It should be dead simple to use with plain RISC-V Assembly files. It has a userspace-only mode and a minimal privileged mode: https://github.com/AntonLydike/riscemu

Adding new instructions is as simple as adding a method to a class, as long as you are reading straight assembly files. If you want to parse ELF-files, you also have to add the opcodes for your command to the decoder (which is currently not very well documented). If you'd like to do that, feel free to open an issue and I'll work you through it.

I also built a minimal kernel, which works on the emulator, but it doesn't support IO and is very rough around the edges in general.

I must say I am extremely disappointed by my FairPhone 3+.

I have had mine now for about two years, and on the hardware-side it's okay. The camera is garbage, even on the model with the better camera (3+). The build quality is okay. I like the ease of ordering new parts, and it disassembles relatively easily (especially compared to all other flagship phones I am aware of).

The thing that makes is borderline unusable for me is the software.

A software update broke caller identification in LTE networks. All incoming calls would be marked as "Unknown number". This is worse than it sounds, as you can not call back if you missed a call, as you don't know who called. The support said it was aware of this issue, and that a patch was "on the way". In the mean time we should just disable LTE. Well, since the 3G network is dead, the only other option is Edge, which is utter garbage. They fixed it after three months(!).

A while later, the automatic display brightness adjustment just broke, randomly going to the lowest brightness setting, and then recovering a couple of seconds/minutes later.

My Fairphone got the android 11 update a month ago (better late then never), but that just broke more things. YouTube playback of anything more than 720p is unwatchable at 10-15fps. I get regular messages that something crashed due to being out of memory (nothing I was doing crashed). I sometimes cannot open the android dropdown menu, I have to really exaggerate the gesture to get any response.

I am seriously considering flashing a custom rom, as I have had less pain and problems with LineageOS in the past.

This is pretty bad, apparently they didn't even let him contact his lawyer: "The actions of the police are also being criticized because Mai was apparently not allowed to contact his lawyer during the house search". That's just straight up illegal.