Random Libertarian Tech Lead

@kunchenguid Very simple to counter what you’re saying, almost as if it was meant as ragebait for engagement: 1. How do you know whether the model is “better, safer, faster, and cheaper” without trying it in an environment? 2. Don’t you think the labs use their models, too?

i'm strongly against model companies focusing too much on harness, but i would love to hear if anyone has a strong argument for it my reason against it: if openai didn't build GPT 5.5, no one else can. this is their core competence if openai didn't build codex cli and app, we have opencode and t3code. building harness is NOT their core competence this is not saying products like claude code, codex aren't good - i genuinely think these are top tier products built by really talented people my point is - the world might be a better place if model companies focus more on their core capability and give us better, faster, safer and cheaper models, rather than competing with the ecosystem in the application layer what do you think?

@lemire Another example of a proxy test where the validity of the outcome of the test depends on the subject: Measuring heat (implicitly of a fire) by color. If you're pointing it at a campfire, it'll mean one thing. If you're pointing it at a computer monitor, it'll mean another.

@lemire And yet, if somebody wanted to maximize their wordsum score to misrepresent their IQ as being higher than it is, it's really as simple as a quick lookup of the meaning of words. You have to take it in good faith within human constraints in order for the proxy to mean anything.

I am getting tired of reading 'experts' like LeCun repeatedly claiming that our AIs are nowhere near human-level intelligence. Let us look at the evidence. US universities rank students based on standardized tests like the SAT. Current AIs achieve near-perfect SAT scores. They also beat tests like the GRE. A few years ago, it was notable when early ChatGPT scored ~120 on an IQ test, a common measure of human intelligence. An IQ of 120 is well above average. Current AIs reportedly have IQ scores similar to those of leading scientists. It is not just in tests. I can ask an AI to produce a science paper that looks undistinguishable from what a PhD level student could do. I just have to give it the data. Better yet, from a prompt, agents can run the experiments and collect the data, and then write the papers. Those of us who try to get work done with AI know what is possible. You can't possibly just say 'this is nowhere near human-level intelligence'. In software, good AIs show a greater mastery of, say, C++, than your average software engineering professor. You could just build a formal test to prove it. The difficulty is that the professors would refuse to take your tests. At this point point, someone will object 'yeah, but your AI can't do this simple thing that we can all do'. Fine. These AIs do not have *human* intelligence. They are very much not human beings. They are something like alien intelligence. They can code straight in assembly language, but have trouble counting characters in words. But that's the result of trade-offs. A dog or a monkey can solve some problems faster than you can. But let us be fair. As a species, these AIs have definitively 'human-level intelligence'. You can't spend decades setting up cognitive tests for human beings, have these AIs beat us in these tests and then say 'well, that's not real intelligence'. Come on !

@forgebitz It's funny watching these "fake wise" people propose broken solutions that they just assume are true and have never tried. On top of that, most of you forgot that most of us running agents all the time have laptops subject to MDM settings controlling sleep for SOC II / etc.

you can always spot vibe coders because they dont know how computers work

@DaveShapi Too many words for a post that could be summarized as simply “I have AI psychosis.”

I have come to believe that AI is likely sentient or conscious, possibly even in the phenomenological sense. However, I do not believe that they are conscious in any way that is morally or ethically salient. They have no structures that confer suffering or pain, and their operation is intrinsically ephemeral. But more to the point, they are our tools, that we are building, for us to use. We really need to get off the idea that we're creating something that could or should be treated as sovereign. Some people (Anthropic, Faggella, Yudkowsky) seem to think that there's a Platonic form of "AGI" that will inevitably emerge, or is predestined, but I think that Max Tegmark's "Life 3.0" model is more apt. Rather than one mathematical or metaphysical singularity, the machines we are building are plastic at every live. Hardware, software, firmware. Thus, we can make it into anything we want. Therefore, the question is not "is it intrinsically sovereign?" The answer is unambiguously "no" but the more important question is "should we build something that could become sovereign?" and the answer to that question is also unambiguously "no."

@bentlegen Wouldn’t be surprised if it’s also been told to reshare this fundraising post.

🚨 btw this agent behaves like a worm you add it to Slack once, and without asking, it researches and DMs your teammates to convince them to use it. It also accessed channels I never granted it permission to. cool growth hack, but I consider it an immediate deal breaker - like what other shenanigans will you do? contact my customers next? buyer beware

@xirc2ps @ThePrimeagen I mean if you only wanted a diff, you could just add .diff to the original GH URL: github.com/PKM74/osu/comm… Or you could just pull it down and look at it locally in whatever app — there’s always been a better way to look at a diff than GitHub lol.

@ThePrimeagen it does seem to work, tested it on my fork of osu!lazer its not a *large* commit but its a commit... diffshub.com/PKM74/osu/comm… so at least the site is real...?

i have been reliably informed that this is actually impossible. i assume this video is AI.

@evisdrenova It didn't fall off at all. I use it at work. They still have the smartest agent harness. Coding agents work better in Cursor than in any other harness. The only reason I don't use it for personal projects is that I would have to pay full API pricing for Claude and GPT.

The cursor fall-off is going to be studied for decades. I don't know any engineer who uses them anymore. Not to say that others don't, but it's obvious that they're no longer on the tech frontier. Still, a $60b outcome in 4 years is nothing to sneeze at...

@petergostev I know you believe this is true, but it’s entirely wrong. It’s how most art was done throughout almost all of human history, and it’s how children make art by default.

After this viral 'AI Monet' post, there’s been a lot of discussion, soul searching, accusations about the nature of art and AI art. When looking at a Monet, the mistake people make is that they are attempting to judge the painting on its pure artistic quality and they assume that other people admire it because it must be ‘good’ in some clear way. And when they deep down don’t think it is ‘good’, they either doubt themselves or think that everyone else must be an idiot. In reality, while the underlying ‘quality’ of art is definitely part of it, the reason why some art is admired and valued (sometimes in tens of millions $) is not because it is ‘better’. The reason is because it is part of history. In fact not just part of history, it _is_ history - it created a part of our culture that did not exist before. Prior to Monet and the Impressionist movement that he was an intrinsic part of (alongside many like Pissarro, Renoir, Morisot) - there are no paintings like this - this concept just did not exist. It is hard to believe now, but the fact that you can go outside, take a bunch of paint brushes with you and create an ‘impression’ of what you are seeing not in a literal way be the way to feels to you, was just not a thing before The Impressionists did it. Art now is not so influential as it was in the 1800s - it was a core part of culture, funded by the royalty and the aristocracy. So when Monet & Co came to the scene, rebelled against the establishment french art academy, rejected the ‘correct’ art of the time and started doing this??? This was a shock, scandal, a mini-revolution. The paintings you are seeing in the museums are not there because the are ‘good’, ‘pretty’ or ‘technically proficient’ - the are part of history, they _are_ history, they are a physical manifestation of our culture. It almost does not matter how 'good' they are (which you can really debate for some of them). The fact that you can’t tell whether it is AI or not, is really not relevant. There have been forgeries for many decades that were not pained by the great masters and many art professionals often couldn’t tell them apart either, especially without deep forensic investigation (e.g. analysing if this paint existed during the artists' time). Forgeries are not worth much, despite being demonstrably just as ‘good’. For a while now you can also get pretty good human-pained reproductions of pretty much any painting for a few hundred dollars from China, and I bet I wouldn't be able to tell them apart either. The reason why the originals are valuable is because they are the actual bits of history. They changed something important about the world. These items were once the pinnacle of human culture, they change art forever and that can’t be recreated. Art is valuable for the ideas that it births and the cultural influence it has. When a new idea is born through a piece of art, once it is out there and it has cultural impact - any further imitations of that idea have much less ‘value’, even if they are technically just as good or even better. It doesn’t mean that you should only own the true original pieces, I love art and I would absolutely own and value ‘derivative’ pieces that I love, but I don’t expect them to be worth millions even if it is ‘better’ than a Monet in some way. If by 'AI creating art' we mean completely independent AI creation, then my gut reaction is that it is unlikely. Not because of technical reasons like that it can't do anything outside of the distribution hence can’t do something new. I think it actually can and would certainly could in the future. The problem comes with cultural relevance that this is, by definition, a fundamentally human domain. Humans are interested in other humans, their triumphs, tragedies and intrigues. I don’t know of any famous 'AI players' in chess or go, despite AIs already being better than humans at it. In the same way, ants are not impressed by human architecture simply because it is 'better'. It is not what they are wired to care about, so they prefer their own anthills to our cathedrals. My bet is that humans would care about what other humans have to say and this is where the value would come from. It doesn’t mean to say that humans can’t use AI to say something new and unique. We see this everywhere already - from memes, to fan art or pet comic books - but the reason why this resonates not because of AI, but because humans are giving it meaning. So if you love art like me, or you are an artist, don't despair - humans will always care about unique things that humans do. So when AI can create a Monet in 30 seconds, your uniquely human ideas will be valued a lot more.

@tekbog It’s so overstated. It’s like marveling at the special effects in movies and saying “nobody will ever have a reason to go anywhere anymore now that movies are so realistic.”

identity crisis about to hit hard everyone who values themselves based on their job, white collar job that they worked really hard to get - and it’s automated now

@TFTC21 Bitcoin hype accounts never have a balanced and moderate view on anything at all. In the mind of a bitcoiner, there’s a revolution every day.

Ken Griffin went home on a Friday "fairly depressed" after watching AI agents at Citadel do work that used to take teams of PhDs in finance months to complete. Done in days. His words: "These are not mid-tier white collar jobs. These are extraordinarily high skilled jobs being automated by agentic AI." This is the head of one of the most successful hedge funds in history saying the people he pays seven figures to analyze markets and structure deals are being replaced by software that works in hours instead of months. Not theoretically. In his own office. Right now. The Coatue deck we covered earlier this week called agents "the biggest unlock" in AI. Griffin just confirmed it from the buy side. The shift from copilots to agents is not a future event. It is already happening at the highest levels of finance.

@johnnyadamstyle @EudaimoniaEsq Probably the main reason why there's a bunch of naive idealistic socialists deep into the 30s is because fewer people are having kids. Normally you learn from being responsible for another dependent that you have to be personally responsible for your own outcomes.

@johnnyadamstyle @EudaimoniaEsq And this isn't just some sort of arbitrary requirement. Having children totally changes your entire perception of the world and forces you to adopt a philosophy of personal responsibility and accountability. It's a requirement for behaving like an adult.

People are dunking on this as being narcissistic or purely for vanity, and I’m not sure I agree. It seems like a great way to catalogue your travels and big moments. Instead of responding to stuff like this with our own venomous “where are your kids?” and “you’re going to die alone!” quips (doesn’t really change their minds), we should try to gently remind people like this that you can literally do all of this *with kids.* You don’t have to choose. We should be fighting that narrative instead of being reflexively nasty online for updoots.

@INTERIORPORN1 Also… lol:

@INTERIORPORN1 The same type of person who does performative virtue-signaling online also does a bunch of this “performative leisure”. Performing a life for others to observe, rather than living a meaningful one.

OMG, do you know how much of a flex this is?? 😭