alex-moon

Ahh AoK is such an amazing game! I kind of want to download it now. I was never very good at it as a kid, spent _way_ too much time building my own campaigns with triggers (this was in the early days before I had discovered programming languages) - obviously no-one ever actually played these.

Really interesting to notice a steadily increasing proliferation of anti-open weights and anti-non-US model rhetoric in the HN comments. Yesterday I saw a comment on another thread claiming a FOSS agentic SWE provider "shill cheap and shitty models", the kind of thing I'd expect from an Anthropic/OpenAI astroturfing campaign or a person with some kind of crippling substance abuse issues doing severe and ongoing brain damage - not really clear which it is but I can't readily imagine a more reasonable third alternative.

I have.

Strictly speaking this is in fact correct and is part of the drive for sovereign AI more broadly including Mistral.

This is an excellent write-up and has successfully changed my opinion on something I think will prove critical in the near future: I always laughed at humanoid robots, assuming they were being built for "comfort" reasons. I now realise this is almost certainly the most reliable way to solve the general robot problem - not just for the reasons the OP lays out, but because the whole world is already built for human-shaped agents anyway.

"You could have invented..."

Oh you know there is gonna be maths in this one. (clicks link)

"Here are the terms we will use throughout: a 'memory store' is a function f..."

Yes. Ha ha! Yes!

I may be missing something so apologies if I'm telling you something you already know but: there are plenty of web search APIs for agents - the Brave one is the one I use: https://brave.com/search/api/

Really disappointing to see this engagement bait on HN. It's all over Reddit now, so I guess it was only a matter of time, but still. I've flagged this and am hoping mods act quickly to stamp this out before it becomes widespread here as well.

I think because we're all just using it thinking we have found the "model for me" and never mentioning it to anyone because what would we say? It's good. It's a bit like the Logitech MX Master, as more and more people assumed they had found the ideal mouse for their purposes, it quietly became the professional standard through sheer adoption.

This is true of all technology. What is special about AI?

> we can fairly say that Hinton was wrong because his prediction did not reflect what a radiologist actually does: had Hinton bothered to ask the question, he would have learned that practitioners do much more than interpret diagnostic images.

This has been the crux, for me, of arguments I have had with non-technical people about why, as a software engineer, I am intensely relaxed about AI "taking human jobs". Recently Ford quietly started re-recruiting engineers after they discovered the hard way that their engineers had been doing more than they thought they had, and I sent articles about it to friends I had previously been discussing the matter with. It is still mind-blowing to me that a business historically famous for automating manual labour can have made this mistake, i.e. of literally not knowing what their staff actually do.

Humans aren't machines, who would have guessed? But, perhaps more importantly, machines aren't humans, and the adherence of a lot of domain experts (that should know better) to fairy tale fantasies about the tech they are literally building genuinely makes me wonder if the whole field has just gotten dumber in the 21st century.

It's said on every one of these but it bears repeating: existing cybercrime legislation already covers this - "rogue agent AI associated with OpenAI attempted to hack xyz" = OpenAI attempted to hack xyz.

A summary of a research article isn't a research article. Do you think newspapers are scholarly journals? Also, you do understand that anyone can just upload a document to SSRN, right? "Preprint" doesn't actually mean anything academically. Also, the methodology of that paper is absurd: Google trained a model and then asked that model what it thought about AI policy. Can you not think of any way bias might be less than controlled for doing it this way?

That's it yeah. These kinds of pieces have been written for centuries. Surprise: people like what they are familiar with! There is no meaningful conclusion to draw from this. Look at Monet or Satie. In their own time? Grotesque, offensive - the kind of art scruffy pompous intellectuals enjoy. Now, a century later? At home in a greige IKEA showroom house being sold on the open market. To be clear: I don't think there is anything wrong with this - I love IKEA! I am buying a greige house as we speak! It is just not an interesting observation, and when writers attempt to use it as an argument against making new things, it's facetious at best, and at worst? Well, several interesting possibilities.

> Why are we commissioning styles that we say we dislike? There are several interesting possibilities.

Like what? A shadowy cabal of liberal elites pulling the strings from the shadows? Stripe (who own Works In Progress) founder Patrick Collision donated to the Trump campaign in 2024. Tiresomely obvious what this "piece" is trying to do. Find a new gimmick imho.

For those reading an article like this for the first time, let me spell out the rest of the argument that hasn't been made explicitly here:

The reason, so the argument goes, that <architecture|poetry|music|art|etc.> is so ugly now, is because it is created by an intellectual elite who are increasingly out of touch with ordinary people. Clearly they don't actually like <concrete|poetry that doesn't rhyme|atonalism|blank coloured squares|etc> so why do they do it? because what they feel, deep down, is guilty about what they are: they want to destroy all traces of the traditional culture they come from because they resent it.

It naturally follows, we are told, that these are the same people who are responsible for increasingly liberal approaches to <immigration|race|gender|sexuality|etc> - and the motivation is the same. The increasing prevalence of <gays|trans people|Muslims|Jews> is the result of the increasing power of a political class defined by their own self-loathing. As these undesirables take up more and more <space|time on TV|positions of power> what we face is something urgent: a crisis, a moral failure of society as a whole - typically the people making this argument use the word "degeneracy" for this - which threatens to destroy everything that made our society what it is today.

Here is the problem with this argument: people don't like things that are new. This isn't a recent phenomenon. Opinion pieces more or less identical to the OP have been written over and over and over again for centuries, and the motivation is always the same: social conservatives see an opportunity to seize power by tapping into something intuitively understood about people generally, which is that the world they remember from their childhood, with all its prejudices, is the one they associate with a feeling of being free of responsibility.

There are only two conclusions to take from articles like this:

1. Who gives a shit?

2. It's time to storm the Capitol.

"The aesthetic is more or a side-effect of using cheaper resources and cheaper methods of production."

This is a really interesting argument and very much not the story I've heard. Can you cite any sources for any of your statements at all?

I'm vibe coding a toy chatbot with OpenCode and GLM 5.3 - I'll open source it once it's meaningfully done (keeping the repo private for now until I am _so so_ sure there are no secrets in it!). This has been an exercise in learning about lots of new AI SWE concepts including some tech (Mastra, LangSmith, DeepInfra) - as well as a chance to try some open weights models. The actual useful output is an iteration loop orchestrator based on GitHub issues/PRs which I have found a lot of fun to use. That'll also be open-sourced.

Benchmaxxing is the word you're reaching for here (though code golf is a perfect name for the symptom). At work I use Codex CLI with GPT 5.4 - it hits a sweet spot: capable enough that it'll do the right thing most of the time, but unsophisticated enough that I can follow along with what it's doing and cut it off and go, "Nope! That's wrong because x, try y instead." For users like me, the threshold for PMF was never "Do I trust it to run unsupervised" but rather "Is supervising it faster than doing it myself" - and as the models get more sophisticated, they actually go backward on this. Most importantly: a bunch of open weights models hit that sweet spot around the same time - I've been having a go with OpenCode and GLM 5.3 on side projects at home and guess what.

"Rocket, coriander, cumin? Somehow I doubt the Babylonians had Mexican herbs and spices." Turns out I was wrong about where these come from! All three are native to the Mediterranean/Middle East and would certainly have been used in this recipe at the time.

Oh right, yeah I agree. It's the kind of thing I can imagine the harness downloading in the background anyway (or instructing the user to do) at some point so I can see the sense in just skipping that bit. Kind of comforting they just do it the same way the rest of us do - if it ain't broke and so on.

Yes, that's what I mean - maybe that didn't come across cos I said "bloats". I work for a business that builds web apps for corporates - LibreOffice is bundled in most of the repos of ours that I've seen, a bit like Imagick is in the broader sector.

Someone may need to explain to me what is wrong with this sorry. Is it because it bloats the app with functionality that some users will never use? I feel like that is fairly usual for productivity/business apps generally.

> A signal only carries information if it is expensive to fake. The male peacock’s mesmerizing tail of evil eyes works precisely because a sickly peacock cannot grow one.

This is by far the most important insight in this write-up and I will be thinking about it for days to come.

A real hacker doesn't hack the software system...

At the place where I was taught this lesson, it was "getting Hasselhoffed" - in short, you find photos of David Hasselhoff in Baywatch (wearing as little as possible, you see? Speedos were very much in in the 80s), and you set it as the desktop. Perfectly harmless, like actual porn would be too far, but it was far enough to be genuinely embarrassing - not least because every Dev gets the same warning on day one (you leave you computer unlocked, you get Hasselhoffed) so everyone knew what it meant anyway. Everyone would, of course, play along - "Didn't realise you liked Hasselhoff that much - bit of an unprofessional image though don't you think?" And so on. If I remember rightly, I literally came in from a cigarette and the first person I saw just went "Do you admire David Hasselhoff's body?" and I already knew what I had done.

This is great but it would be useful to have some indication of cardinal directions - it is not clear to me which way I am facing.

I am struck by the resemblance between this logic and the logic behind drinking. I remember back in my 20s one of the devs would come over at the end of the day and go, "Coming to the pub?" The standard answer was, of course, "I might come for one." The word "one" was a kind of punchline all on its own of course, a running joke.