titularcomment

lol. Simple case of if you have to ask, you probably shouldn't.

1. It depends on model and tokens spent 2. Models talk the talk but not always walk the walk

Don't play fair in a game others cheat.

I'd respectfully advise you to assume less

GPU driver support is not limited to CUDA. Devices that require binary blobs of firmware to work properly are not always exotic. Some examples off the top of my head that won't work with OpenBSD hassle-free is Vantage (LLL), Solaar ... but I haven't tested. I understand you finding it easier to work with and it's faster install speeds do pique my interest, yet you being on HN already puts your technical literacy on some low % of gen pop. OpenBSD may be comparable to distros like Gentoo, Void or Arch but certainly not out-of-the-box ones like Ubuntu. They serve different auidences.

Personal laptop use with Ada Lovelace or Ampere family NVIDIA GPUs (did you mean this or CUDA, i couldnt tell), personal desktop use with unusual peripherals, dependency on ports, existence and competition of FreeBSD etc. I love OpenBSD's code philosophy (they were the first to introduce a lot of security techniques[1]) and the programs they produce, OpenSSH is a lifesaver, and I use doas for its low footprint on my Linux machine. Still, they have a convulated install process, dubious hardware and software compatibility that is better solved in the Linux world today.

[1] https://www.openbsd.org/innovations.html

I believe issues will arive with AI in the near future in terms of their performance. Disagreeing with the author, I think code quality pre-AI, in certain languages, were golden. They all contained bugs but they were written in a way that bug sources would be more or less obvious and most but not all top open source software had linting guidelines, code styles & friends to help make it ingestible. Perhaps these helped said project's maintainers a great lot, but it absolutely SAVED AI when it came to training data. These top-quality repositories with thousands of lines of great code combined with hundred thousand lines of mixed quality code was the perfect formula for the big data churner™. Now what? All those repos, inflated with stars are generated by one of the 5 LLMs out there. There has been extended discussions about deterministic behaviour in LLMs, and I'm no ML engineer but to me this drop in entropy will surely cause backtracking in code quality. Of course there a vast array of improvements that can be made outside of training data, but the whole psychology attached to LLM marketing, in my opinion, obstructs those improvements. You could train a better suited way of input for LLMs, set cutoffs and posttraining in just the right places and everything but you gotta think it through. How many of those who ask their favourite LLM everything think anymore? How much of our current knowledge is safe from AI hallucinations or subtle nudges? AI has changed the world of coding, but the current state in Anthropic HQ will determine if it will just be mid-quality codegen or accelerationist fever dream.

I don't understand caring about 'slop' in creative media. Its pressingly more important in programming because there are a lot of bugs linters, complicated language models and even fuzzy sandbox searches can't catch. Even if they did, coding is collaborative work and maintaining vibecoded infra is a genuine hassle. However, none of these apply to music or books. AI or not, humane feelings always shapes a piece to its last form and its entirely up to you to determine if you like it or not. AI generally fails to capture this, and if it didn't - a fluke -, why not enjoy the piece?

Yet a good enough reader can distinguish between appearance and content. Better yet, everyone has their own type of favourite content. Maybe someone out there enjoys a 300 page Claude essay, you never know. Therefore the 'books after 2023?' argument is senseless sans the purchase part in which the contents may not be immediately obvious.

OpenBSD is the Linux of a decade or two ago, not attracting attention and not being compatible or useful for quite a lot of stuff.

If an exploit is actually working, human effort is void and the ML has done a great job. However most of the time its hallucinating, confidently talking gibberish in technical lingo. This phenomena is only amplified by those who try to make a quick buck without effort using older models, not reviewing output or prompting properly ('find me bugs in Linux, make no mistakes')

do you truly think that every LOC on those services were generated by AI, lest not reviewed?

I'm glad that you asserted that you do have potential hidden bias and say that you used an LLM to judge the tone of an article partly pertaining to AI usage in coding right after.

The main difference is (simple) calculators are deterministic and monotonic. Meaning it executes a set of instructions in a predetermined way to produce its output. Bringing LLMs to that level is a whole another ball game. But we'll see, perhaps the arithmetic nature of algorithms will be replaced by a whole lot of tensors in the near future.

Yet another post and forget firehose app

Painful read.

Related interesting project: https://github.com/EtherDream/web2img

I believe poisonous loops of non-sense text are the best choice in terms of LLM capabilities and human distinguishing potential; the next iteration could be non-sense with reasonably intact grammar and content. At the very least, show some content whilst doing the PoW (isn't the point rising computational costs? Give me something like YouTube video decryption instead of having me wait at least) OR use the PoW for some useful protein-folding, finding the next prime, or an alternative monetization scheme.

My guess is its an implementation error, not an hardware limitation. I have two 10-year-old devices and one passes instantaneously while the other halts for a good half minute every time.

Abuse such as this wasn't uncommon before, email platforms with lax ratelimits have always been abused through their clients' unsecured infrastructure. The only difference in post-LLM world is the amount of platforms as well as clients popping up in this space with dubious code quality that may lead to more attacks as;

a) having an email-sending product typically meant you had a project with a lot of effort invested into it as well as knowledge

b) the models, tokens spent and review done differs in the world of vibecoding and there is a race to the bottom to produce, produce, produce. Quantity > quality

Dots and periods. Everywhere. So many. There is no paragraph — its sentences all the way down.

That made me think if the project is entirely vibecoded as well.

Even for a project manager without network access, hosting flawed software on your LAN can only get you so far.

I believe new (programming) languages will emerge both for LLMs to parse and take instructions from as well as for them to generate code in. The former is because English is a nuanced language evolved for human usage which the LLMs don't quite need, with the only advantage of it being a metric ton of training material. Same goes for Rust, Go and other languages LLMs do primarily well coding in, which all have concepts geared towards human convenience.

I agree. From the get-go, Bun was apparent in its design philosophy: we do everything you'd ever want; runtime, bundler, test suite, package manager, all in a new breaking patch each week. With each and every one blowing the established competition out, better, faster and stronger. But it was glaringly obvious that they'd do anything but Keep It Simple Stupid. It was obvious that the only production environment it would see the light of the day in the near future would be YC startups burning one after another at the speed of an accelerant. Now, they're past the point of no return.

the fact that this works, as well as cloudflare having a literal web scraping tool available as another product honestly makes my blood boil.

Asking out of curiosity, have you had material success with this technique?

At least an AI has a chance to ingest it all

Solving humanity's issues, one more issue at a time

This particular example is not contradictory if you think about it, its essentially a protection racket.

Maintaining software is like 80% of the job.

Obscurity can be combined with security for much better results. Machines as well as AI in extension thrives off patterns and making illogical off-pattern decisions is usually to the benefit of the defender, not the attacker. As you said, the attacker has a wide attack surface to cover while the defender only has to fortify his home NAT. E.g. port knocking may very well throw off the hoard of scanners on the wide net simply because its not standart and the combination is known only to you. Similarly, fail2ban may not work as well on a standart SSH port, because every attacker is going to hammer that and one may get your misconfigured password root login right.

Now, in both instances, the obscurity provided does not necessarily cure your infrastructure's vulnerabilities, a dedicated attacker wouldn't have a single problem with either of these. But for someone who hammers the whole internet in a dim hope of finding another Wordpress server from 2017, or the latest flawed online security cam, your disguise is as good as perfect.