Like a lot of folks, the 'tells' of LLM writing for me are more of an 'I know it when I see it' than a specific set of hard criteria (which is probably why it's so hard to make a tool for it), but if I could sum it up, it's something like this:
I always had a high faculty for language, and when I was young, even very young, I was able to speak in full sentences much like I'm doing now. What I didn't have was context. I didn't know how to relate to people socially, and then later on, I wasn't as experienced an engineer as my peers, but I was trying to act adult and mature in both ways.
So I tried to write in ways that were catchy and unnecessarily complex, because I thought that conveyed sophistication and intelligence, and that I knew what I was on about more than I did. It's the kind of writing that gets you through a 30-page final paper for a class you don't really like, or a research project you didn't finish. The term I often use when complaining about my colleagues' LLM output is "flashy but meaningless."
As my confidence grew, I started writing in simpler (though still not the most concise, that's just not who I am!) ways, because it became easier to convey ideas simply in a way that made it obvious I understood them.
When you think about what LLMs are - word machines without (or with very limited) context - you can see why they write the way they do. They are always acting; trying to replicate a human contextual experience with only weights and tokens to carry them through it. That fact is part of why I am skeptical of their ability to suprass humans in reasoning based on their own output.
The Switch is powerful enough it could just straight emulate an N64 game if it needed to. It has a huge back catalog of games where it does this. It's basically a computer on the level of a mobile phone or tablet, or a laptop from 15 years ago.
The 3DS is still a _console_, purpose built and limited in what it can do. It is also approachably easy to find a good one in working order, and still has the vibe of a retro console.
I think the API pricing is a good proxy for "what was the actual physical cost of doing this," but even if you increase it from 5k to 30k or even 100k, it's a huge savings over getting a human to do it, AND it was faster.
I don't think the productivity gains from AI can be ignored for contexts like this. It's what they're best at.
I just hope it goes to good uses (more ports of fun games, more interoperability and opennesS) instead of bad ones (firing humans instead of retasking them), and that hope feels faint.
Yep, I agree. Further in the notes it's mentioned this probably would've taken a handful of Senior Engineers anywhere from 6 months to 2 years. The range of budget on the project is like $200 to $500k. This was done for peanuts in my spare time. I share both your optimism and pessimism, depending on the day, lol.
In no way am I suggesting that doing this without AI would be more practical. What I want to know is roughly how much is Anthropic subsidizing this stuff through the subscriptions currently and how much we can expect the real price paid to go up in the future.
That's not true, companies pay humanies to develop ports and remasters of software all the time. Maybe not from one dead platform to another, but the same technology can be used to take a game like this from a dead platform to a living one.
It was, 15+ years ago, when porting N64 games to 3DS was a lucrative market for hired guns. We have the magic 'port' button and this tech just showed how easy it is with severely different architectures & constraints.
I refurbish these old Thinkpads sometimes. They're tough to work with but I love the little flapping duck and the whimsical BIOS design. A simpler program from a simpler era.
I sell a lot on eBay and this is reminiscent of the experiences I have with the worst eBay customers. There are folks out there who will just push and push and push for whatever it is they want; they're the reason policies exist, and they ruin it for everyone else.
There's an inherent tension between the person providing the service and the person consuming it that has to do with fairness and leverage; each person always feels they are on the wrong side of the deal. In reality, it is the aribtrator, who sets the rules, who gets the upside of both deals, so they're right, ish, but they blame each othner.
Airbnb is in many ways reinventing hotels from first principles. The economics and incentive structure of "just let somebody crash on your couch" are not at all the same as what Airbnb is currently doing or expecting guests/hosts to do. Guests treat the place like a hotel, but don't act as respectfully as they would in a hotel. Hosts charge what it costs to run something like a hotel, but don't provide the level of service necessary to qualify as a hotel, and at the same prices. Everyone is unhappy because everybody's expectations are unreasonable.
Who is creating the perception of what the product is? Airbnb. Who is making decisions that piss off both groups? Airbnb. But they're also squeezed by the unrealistic expectations of both sides.
I don't see a good way out of this, and the comment section isn't encouraging. People are either "hosts" or "guests" and they only care about the side they relate to.
AirBNB has been abused as a short term rental platform by more or less slum lords. This has exacerbated housing situations all over the world, I saw it in Berlin. Portugal has had similar problems.
It is not a serious platform IMO. Just like über, it invests absolutely nothing in the infrastructure it feeds off.
Airbnb is consumers learning short term lodging regulations over again. Crypto is consumers learning financial regulations over again. Ridesharing and delivery gig work is consumers learning labor regulations over again. It's all just an economic grift to capture the duration between when the starting gun fires and existing regulations apply.
This is one of the most artfully positioned, full-throated, thoughtful, and capable takedowns of the AI hype (not necessarily LLMs as tools) I've read. It doesn't try to speak doom to AI companies, it doesn't suggest the tools are worthless at things they're good at. Instead, it disassembles the assumptions underlying ideas about frontier AI labs' promises (of both utopia and doom) by explaining in simple terms what the real accomplishment they've made is:
They are thieves who turned their stolen goods into a pretty neat invention.
They are not prophets of the end times or messianic figures bringing in a new age of human relaxation or the biggest trillionaires since Tim Cook and Elon Musk. They are people with one good tool that they think can and should be used to do everything. They won't accept that their tool is just useful, or helpful. They want it to be _more important than the humans and human work it was based on_, and nothing less will do. They want to replace effort and thought with simulacra of both and be applauded for it.
Lest this go down as an anti-AI post: I use AI tools. I use LLMs every day. They are useful... but they can't do on their own what we did not demonstrate how to do first, because they are tools.
I could be wrong, but I don't know that I see this changing.
> but they can't do on their own what we did not demonstrate how to do first, because they are tools.
Funny thing to say here, where an agent swarm (1M hours agent time) just did something no human did. If someone else solved NS blowup before OpenAI we would have heard their claims already. And we did hear partial claims, "they rushed in the same direction" kind of claims, not "they copied my fully developed idea". I think this is Move 37 moment for math, when we realize we might not be the top anymore.
> How could I be wowed that a Millenium Prize problem has been solved using Lean verification for its proof in 2026, when that was on our minds for our 5-year plan back in 2021? How can I be impressed that OpenAI claims to have solved it in 88 hours when even I, someone barely involved, spent more hours than that working on the tech they're using?
> How much work from how many people did it really take to get that result?
I've watched this creator since, I don't know, 2010. They seem to be a very creative mathematics educator. This the point that is being insinuated here is ludicrous.
It's like saying, "how can I be impressed by copernicus' heliocentric model, when he built it on the 500 years of painstakingly collected data from the babylonians?"
It's either disingenuous, or ignorance.
> This should be embarrassing enough for OpenAI even if they didn’t have access to Buckmaster and Alpöge’s ongoing work.
More repeating of unsubstantiated accusations fanned by the ongoing culture war.
The rest... Gee whiz. It just hits all the notes. "Anyone who disagrees with me is a pedophile and has ai psychosis."
This is a diatribe from a person who makes wonderful art videos geared towards math education, against an organization solving millennium problems.
That is very much the case at the current iteration of AI, but that should change with the introduction of world-mapping, computer vision, and world interacting AI. If machines are built out so that they are constructing accurate world-maps of how the physical world works and how to operate in it, then AI will begin to be able to out-invent human tool concepts and push past the horizon of what humans have so far been able to dream
I think that's the wrong angle with machine learning.
The biggest limits on machine learning are how many examples it takes to get good at something. Right now, this is at a literally evolutionary pace: our bodies/genomes have barely adapted to the invention of controlled fire, thousands of generations ago, yet if it took thousands of examples for our brains to learn then we'd sill be unable to control fire.
Computers only partially make up for being slow in the "by number of examples needed" sense by being fast in the "wall clock" sense: If a biological brain took as many examples to learn a thing as machine learning takes, the creature would starve before figuring out how to eat; yet transistors operate faster than synapses by the same magnitude to which a jogger outpaces continental drift.
The "only partially" is why, despite around a trillion miles of driving experience across the fleet, you still can't buy a Tesla without a steering wheel (and even the Cybercab that's supposed to be without a steering wheel is often photographed with one).
> or input generated by them as derivatives of human input
Trained mostly on their own data, not ours. The mix changed, from 10% to 90% synthetic data. RL training is now the largest part of the process. The AI work is searching, and I don't think we can claim this process to be credited to us anymore. The origin point is, but not the journey.
This is a very basic misunderstanding of machine learning. The goal is always to learn a representation that generalizes beyond the training set without overfitting. If you succeed in generalizing, you can do more than produce derivative work.
That’s the thing. AI won’t be limited to being trained on only human input, why would it? Humans are trained by interacting with their environment plus a genetic world-mapping initial seed. There is no reason AI will not be the same, just at a much more extreme level
It depends on the knowledge domain. I would argue that much of human knowledge is individual to the person acquired by interacting with their environment
This isn't a bad idea, but please, I am begging everyone here. No more AI prose on your websites.
Please take the time to read and edit. Get rid of the emdashes and the pithy little "The quick start lands here. Nothing expires into a paywall." / "Sign up and pay in one flow — checkout follows registration." zingers that are meaningless or already implied.
Please stop allowing pragraphs like this:
> Bounds are per organization — one organization can hold several teams, and its message count is metered across all of them. The monthly allowance is the real limit; the burst ceiling exists only to stop runaway loops and sits far above a normal day’s work.
to appear on pages where you expect humans to read them. They are worse than useless to read.
Obvious AI copy on the landing page says to me that you don't care about the product.
I know the thing was written by an AI. That's cool: I do the same myself. They're better at code than we are.
But the product is the thing I'm buying. It's the reason I would consider giving you $50/month rather than getting Claude to knock me up my own version. If I don't think you care about the product enough to explain it in your own words, why on earth would I ever give you money?
(Not a personal attack. I feel this way about every new site and it seems like you're interested in feedback. I wish you all the best.)
I hear you loud and clear, thank you. I need to prioritize fixing the copy.
(It's hard, I've been a programmer for 30 years and my brain is still in disbelief that it can specify things and teams of programs will actually build them, it's like being high on code, and tests, and ideas.)
Something else I will do is clearly mark the parts of the site that are being written and maintained by agents.
Incidentally, I do actually care about human writing. Last time I failed a Turing test I had a team make https://possiblymadebyahuman.com just for fun.
You are probably still writing for agents. It's more than likely these constant "AI smell" insults on HN are simplY themselves agents scanning posts for any AI smell simply because it gets upvotes to post such slanders.
We KNOW that people don't like reading AI's writing. Even the AI performs better with quality content. Why not just... edit the text?
Did the creator even read the text? If they put real effort into the product, they would have also probably cared about how that product is presented. This shows me that they don't, so I'm not going to read it, and I'm certainly not going to buy it.
I always had a high faculty for language, and when I was young, even very young, I was able to speak in full sentences much like I'm doing now. What I didn't have was context. I didn't know how to relate to people socially, and then later on, I wasn't as experienced an engineer as my peers, but I was trying to act adult and mature in both ways.
So I tried to write in ways that were catchy and unnecessarily complex, because I thought that conveyed sophistication and intelligence, and that I knew what I was on about more than I did. It's the kind of writing that gets you through a 30-page final paper for a class you don't really like, or a research project you didn't finish. The term I often use when complaining about my colleagues' LLM output is "flashy but meaningless."
As my confidence grew, I started writing in simpler (though still not the most concise, that's just not who I am!) ways, because it became easier to convey ideas simply in a way that made it obvious I understood them.
When you think about what LLMs are - word machines without (or with very limited) context - you can see why they write the way they do. They are always acting; trying to replicate a human contextual experience with only weights and tokens to carry them through it. That fact is part of why I am skeptical of their ability to suprass humans in reasoning based on their own output.
reply