MostlyStable

Sure, that's fair. I enjoy the process of figuring out ways software can solve some (usually quite trivial) problem for myself, but then I'm very happy to hand the act of making it to a robot.

Exactly. I have a backyard garden not because I need the food (it would be infinitely cheaper and easier to buy even high end farm-stand vegetables), but because I enjoy gardening. Buying a gardening robot (if such a thing existed) would completely defeat the point.

I use an LLM to make custom software for myself not because I enjoy coding but because I want software that I can't afford to hire a human to make for me. Buying a robot to do it (current agentic coding LLMs) is the only way it gets done at all.

These two things are not the same, and should not be viewed in the same light.

All that being said, I would 100% buy a weeding robot. Weeding sucks.

From a mathematician who was intimately familiar with some of these problems [0]

>I don’t understand it yet. Maybe it’ll take me an afternoon to check all the calculations, but what would still be missing is why this was an approach that would’ve made sense in the first place. Is there some broader context or theory within which this would’ve been the obvious thing to do? What other results can be proven using these techniques? What is it telling us about quantum information or operator theory? I have no idea. I spent about an hour this morning asking ChatGPT these questions, but it’s somewhat frustrating because it speaks with a mishmash of physicist, operator algebraist, quantum information theorist-lingo, plus the usual LLM breezy lilt that annoys everybody.

They certainly seem to have "intuited", in a way that is not immediately obvious to experts in the field, the way to solve at least some of these problems. This was not just simply grinding away at a method that humans already knew would work and just hadn't gotten to yet.

[0] https://nitter.poast.org/henryquantum/status/208362369543662...

There are loads of unknown (to science) deep sea sharks out there, most of them quite small. A friend of mine in grad school spent several trips on a deep sea trawler out of Madagascar and described multiple new species.

One of the reasons why I loved The Food Lab was that very often Kenji would write several versions of a recipe at different effort levels, along with detailed reasonings about why and how he shortened a step and what one was giving up by doing so. While he didn't do that for the Pho Ga recipe, he did do that for his Beef Pho recipe, with a full blown 6 hour "traditional" version [0] and a faster (although still not as fast as the pressure cooker pho ga one) 1 hour version of the recipe [1]. My guess is that the write ups of the 3 recipes together probably give one enough information to dial in relatively precisely how much effort you want to give and which steps one can skip or reduce.

[0] https://www.seriouseats.com/traditional-beef-pho-recipe

[1] https://www.seriouseats.com/quick-and-easy-1-hour-pho

Interesting to see this posted, as it's quite an old recipe (the oldest capture I could find on the Wayback Machine was from 2015 although I'm pretty confident it's older than that; as an aside, can I say how annoying it is when websites only have an "updated on" or "edited on" date and not a date for when it was originally written?). It's also one of my favorites of Kenji's though and a regular in our home.

I think that Kenji's series of pressure cooker recipes are probably my favorites of everything he did when he was doing The Food Lab section. Almost every recipe is fantastic and they are mostly pretty easy. If I had to pick my single favorite, it would probably be the pressure cooker chicken enchilads [0], although that one is on the higher end of the effort spectrum

[0] https://www.seriouseats.com/pressure-cooker-fast-and-easy-ch...

Given that there is already a simulation, with very clear performance metrics, I wonder if anyone has ever tried making an evolutionary algorithm and if so, how well it would perform. Do we have any idea of how a theoretically perfect algorithm would perform? How much theoretical performance is there to be picked up?

Yes, I agree that this is still a datapoint for real world utility of AI assisted vulnerability fixing. But it's not as simple as an apples-to-apples comparison with previous years.

I have recently noticed that the words "ad" or "marketing" have become, in and of themselves, with no additional information or context, slurs or dismissals.

I understand why. The modern internet has turned advertising into a morass of constant bombardment and the only sane response is to block as much as possible and ignore as much else as possible.

But it's unfortunate because, in some sense, ever single thing that a company every says that is not legally mandated in some way is a form of advertising.

And in many cases, that "advertising" contains true, useful information that can be helpful.

What is important isn't whether or not something is an "ad", but instead, whether or not it contains true information that is helpful in some way.

Many ads don't reach this bar. They are either misleading, straight up lying, or information that is almost completely useless.

But when I'm searching for a particular product, about the only source of information at all is some form of advertising, and I almost always find at least some amount of it to be helpful in making a product decision.

Ads are more often than not polluting to the informational ecosystem, but that's not because they are ads.

The entire arc of human progress is taking things that used to be the domain of only the privileged and making them the domain of the masses. There was a time when if you were not actively contributing to the production of your own food (to the exclusion of almost any other activity), you were extremely privileged. Now, ~ no one (in the West) has to grow/raise/produce their own food and the proportion of most western's productive output that goes to purchasing food is a tiny fraction of overall work.

A new category of task starting to move along this trajectory is to be celebrated. Maybe some day people will clean some small portion of their house the same way that some people now have backyard gardens.

After doing a bit more research, this fact is somewhat less impressive. Apparently, this was also the first time in nearly 20 years that there was such a large capacity crunch and many researchers weren't able to get into the competition. One of the rejected researchers did apparently have a working exploit, which, upon not getting into the event, they responsibly disclosed, and it was then patched before the event.

I'm honestly unsure if this is a Poe's law thing or not. I'm going to go ahead assume that you are doing the honorable thing of purposefully not including a /s for the integrity of the joke.

I just did a search and apparently this fact (the specific one about no payouts for the first time in almost 20 years) has not gotten a discussion on HN. Given the degree of skepticism around the utility of AI bug finding and fixing (this very thread is full of it), I would have thought that concrete evidence that it can help actually make real software more secure against attacks would have gotten a write-up somewhere.

This is one of those "tell me you don't have kids without telling me" situations. The process of teaching a kid to be a responsible human is long and arduous. If you expect a 2 and a half year old to successfully be picking up their toys, well, I don't know what to tell you. Go spend some time in a daycare, I guess.

And of course, that's completely ignoring the fact that I also would like to not have to pick up my shit if I could buy a robot that would do it. Yes, I can. And in fact I currently do...mostly. But I'm not happy about it and if there was an even quasi reasonable amount of money I could pay that would mean I would never have to do it again, I 100% would do so.

And if you can honestly say that you wouldn't...well I hope you are enjoying your life in the monastery.

These are the kinds of products I now just straight up refuse to buy.

My understanding from parents and friends who have a cleaning service...it pretty much depends on a certain baseline level of cleanliness. Most cleaning services don't (and likely wouldn't be able to provide) basic "pick up around the house". They will come into a house that is mostly picked up and sweep, dust, mop, vacuum, etc. but they won't come into a toddler house and put everything away.

I'm sure that there are services that do this, but I very much doubt they are the $400 per month services you are talking about. Robots may not be able to do that level of task at all right now, but they will get there.

And of course, that's not even thinking about all the cleaning tasks that would be avoided in between the every-other-week visits.

Having my surfaces be somewhat cleaner every two weeks is worth ~nothing to me. It's by far the easiest and least unpleasant part of cleaning a house. Having a robot that will pick up the detritus of everyday life and put it away, pick up and clean dishes, etc. and do so every single day would be worth an enormous amount of money.

I have no idea how long it will take for robots to get to that point, but whichever company cracks that level first is going to make a lot of money.

It is entirely possible to believe multiple things at once:

1. The drop is fertility is bad for society

2. The drop in teen pregnancy is good, despite it's contributions to 1

3. We should try to fix 1. without undoing 2.

Again: yes, very many people (just look through this very thread) are saying that. Tons and tons of people are saying some combination of "the whole thing is made up/a lie" or "This is entirely down to OpenAI incompetence and doesn't matter", or for some other reason (often not stated) making it very obvious that they do not think that this story matters very much.

And, again: to the people who aren't saying that: whatever argument is being made, it seems to me like it probably doesn't need to rely on claiming anything about the motivation of OpenAI. It sounds like you are against government regulation of AI. That's a position that a totally reasonable person could have. You should be able to argue that this event does not justify some particular kind of government regulation without reference to OpenAIs motivation for reporting the story.

Firstly: yes, very many people are saying this.

Secondly: to the people who aren't saying it....then why are you bringing up marketing at all? If the model is capable of it, then the motivation for why OpenAI is talking about it/reporting on it is completely beside the point. Either the capability matters or it doesn't. If the capability doesn't matter, or doesn't matter in the way that some particular person is claiming, then say that and explain why. Just saying "marketing stunt" adds zero value to the discussion.

I'm very open to arguments about why we shouldn't be concerned about this event (although my prior is very much that we should be, not so much because of the capabilities themselves, but because of how fundamentally misaligned this demonstrates the models are), but I am completely over listening to anyone who has nothing to add other than "marketing stunt".

I guess you are one of today's lucky 10,000 [0]

[0] https://xkcd.com/1053/

As they explicitly say, Opus 5 is ~ equally capable as Mythos/Fable at finding vulnerabilities, but it is much less capable at exploiting those vulnerabilities on it's own. That is an extremely meaningful difference and to me completely explains the difference in tone, release style etc.

Just to add on to other comments: reasonable people can disagree about the degree of safety concern with near to medium term AI. But to not address the arguments at all is, in my opinion, a serious mark against the value of this article.

Your answers here are part of why I don't understand this entire viewpoint. You sound like a conspiracy theorist, talking in innuendo, and unwilling to say what you believe clearly.

I'm not quite sure what you mean by that, but it sounds like you're suggesting the lying about it.

All the people saying that this is pure marketing: Do you think that they are literally lying about what happened, or do you just think that what happened doesn't matter in any sense whatsoever, and that therefore the only reason they are telling people about it is a marketing purpose?

I will leave the following since so far all the commenters seem to have either not read the article at all, or to have at least skipped the following two sections:

>The third, predictable and annoying argument is “stop overthinking it, it is just a game.” Which is – if you will permit me to paraphrase C.S. Lewis – an unserious ‘boy’s philosophy’ unfit for adults. Video games are cultural products, like books or films or any other – and to say that those cultural products have no impact or importance is to argue against the historically obvious, given how frequently we find such things at the roots of important political ideologies, social movements, and wars.

>I bring this up not because I demand utter historical accuracy in my games – I will almost certainly happily play Age of Empires IV despite this. Instead, I bring it up because I find that this overly simplified view – sometimes, depressing, held to be the truth by some – in its errors, reveals some important things about empires, which again, I should note, seem to have been the dominant very-large-scale organization system for humans generally in the historical period.

First it tends to obscure the necessarily diverse nature of empires – as we have just laid out and perhaps more to the point, it obscures the power that empires derive from being diverse....

There have been some pretty egregious misuses of eminent domain in the past, but using it for infrastructure is pretty much the quintessential justification. That's basically the only reason to have eminent domain at all. Arguing against it when that's how it's being used is, in my opinion, essentially arguing against the very concept of eminent domain. Which one can totally choose to do! But they should be up front about it, if that's what's going on.

As for the title, I would go further than you do. This is not a misleading title. It is a lying title. I'm usually a pretty big believer in Scott Alexander's Bounded Distrust [0] concept for the media where they very rarely actually, technically lie. This title (although not the article itself), in my opinion crosses that line.

[0] https://www.astralcodexten.com/p/bounded-distrust

About a week ago, I set it up on my windows machine (host) to my ubuntu laptop (client) with the help of an LLM in about 30 minutes, with an additional 15 of troubleshooting the first time I tried using it away from the LAN. It was pretty painless and worked right off the bat. And then it turned out that A) the game my friend and I were trying to play had only local multiplayer (so my host machine couldn't connect) and B) despite protonDB claiming limited compatibility, the game launched immediately with no problems natively on my laptop, so the whole exercise ended up not being necessary. But it was relatively painless.

In case it makes a difference, I was also using Tailscale.

yes, but the GarminDB [0] project still manages to get most available data by impersonating a browser (admittedly that could be broken eventually). If that happens in the future, I'll probably revisit making an entire tracking app for myself.

[0] https://github.com/tcgoetz/GarminDB

I almost decided to make my own app recently (and was also considering the Pebble) to try and replicate the old basis Peak functionality, but when I found out that the Garmin health data was relatively open and accessible, l decided that it was easier to just make a new badge/gamification system for the Garmin data than to make an entire tracking app.