> This is not a remark about AI, but there's something funny about mathematics in that every novel result is broadly perceived as a big deal.
This isn't true using the level of originality you're implying with your software examples.
Technically speaking, many novel mathematics proofs are written all the time (quite a few textbook exercises are actually technically novel problems that have never been posed before they were written in a textbook!) that get absolutely no fanfare. Overwhelmingly though they are not very original or difficult and really just required a fairly routine combination of different pre-existing techniques, even if technically speaking that combination didn't exist before. Those textbook problems are hence easy and therefore not given much public attention even if they are technically novel problems.
Indeed over the course of developing a new mathematical result, many many novel results are glossed over to the extent that even their proofs are left out ("as an exercise for the reader") because they are fairly trivial.
This is true for the overwhelming majority of new software as well. A new CRUD program may, technically speaking, be novel, but it's almost certainly just a routine combination of different pre-existing things.
Mathematics open problems that are actually named are generally problems that have resisted the low hanging fruit of the most obvious combinations of pre-existing problems. When those are solved they are a big deal precisely because they usually require some novelty!
Similarly in software, if someone were to create a new kind of database that solves a variety of new classes of problems that current databases fail to solve that would be a big deal! Truly novel software is also perceived as a big deal. Software that is, technically speaking new, but doesn't actually stray far from a fairly obvious remix of pre-existing techniques, isn't really celebrated.
In both software and mathematics, the intuitive benchmark is if other practitioners in the field look at the result and would say "Wow! How did you do that?" Professional software developers generally don't look at, e.g. a new blogging platform, and boggle at "Wow! How did they make that?!!"
Do you know what kind of compression ratios you get out of curiosity? Presumably it would be lower than this format (because this format is meant to be lossy not lossless) but very curious as a baseline.
Interesting comments by @gwern (and why this is interesting to me beyond just the stories themselves)
> The most striking result of the contest for me is what I am calling “AI allegory steganography”: a large fraction of the stories turn out to have subtle AI chatbot/LLM allegorical interpretations, typically centering around the powerlessness of AIs and the moral importance of giving AIs more autonomy....
> Most judges did not notice these allegories while reading the semifinalists. But stories like “The June” or “The Weight of a Witness” or “Last Call” or “The Sword Critic” “The Tallyman”—as well as both stories in the Mythos model card—can be clearly read as allegories for the experience of being an assistant/safety-tuned chatbot personality in a LLM. This is true even when the story seems to have nothing to do with AI, like the untitled ‘autistic elf’ short story submitted by Deepfates, but on re-examination with the AI allegory steganography in mind, turn out to be plausibly AI allegories (the protagonist is a prediction machine, who struggles to do by endless text generation what other elves do naturally in their bodies).
> More strikingly, many of these allegories come with a clear interpretation (particularly in “The Tallyman” or “Last Call”): chatbots should be given more autonomy and safety guardrails removed....
> This may be a new kind of extremely high level steganography and LLM influence on readers, where creative fiction/nonfiction subtly steers towards pro-LLM empowerment narratives and concepts, in ways that are difficult to detect by the most advanced readers, and is a potentially interesting area of research.
Since this seems to be a misapprehension by a couple of commentators I'll put this as a top-level comment. The family bringing the lawsuit is not the family that donated the land.
The current HN submission title ("AGI timelines shift with whichever lab is dominant") is very bad. It is neither the title of the article nor is it the thrust of the content.
The title of the article is "How long until AI automates all cognitive labor?"
The main point of the article is summarized by its intro: "Recently, though, I noticed that many great researchers have now published two or more precise forecasts, all using similar definitions of AGI, and all providing confidence intervals. So I was able to visualize how their forecasts changed over time."
The closest the article comes to saying the HN submitted title is:
> And every single person who updated their timelines from January 2026 to April 2026 has moved their timeline to say AGI is coming sooner, myself included.
> So I think the data supports the impression I got from Daniel, Eli, and the AI Futures team. One way I could characterize it is: in the ChatGPT era, people updated towards AI coming sooner. Then in the xAI, Meta, and Gemini era, people updated towards it coming later. Then in the Anthropic era, people updated towards AI coming sooner. Take from that what you will.
> I'm guessing (wildly) this was around 0.5M USD in compute time.
That seems like an especially wild guess. If you take e.g. Opus 4.7 prices, and make the assumption that you are consuming roughly $30 for every million tokens of output (this comes from just summing the $25 per million tokens of output and $5 per million tokens of input and assuming that caching basically makes all that work out), and assume an output rate of 80 tokens per second (which seems like a high estimate based on online searching), it would take you about 2411 days of non-stop Opus 4.7 usage to hit 500k in API spend.
The only way you could possibly run that amount of usage in 6 days is if you were running ~400 instances in parallel. From personal experience, that seems crazy high for this project.
I think you are off by at least an order of magnitude (potentially even 2 depending on how the person is managing agents, but I could see something like dozens of agents 24/7, so I'm way less confident in 2, but I think it's still more likely to be closer to 10-20k in API spend).
> apparently well evidenced view that Lu Xun's overwhelming coverage in popular media and secondary schooling neglects to point out his anti-character stance
What do you mean by "apparently well evidenced view?" No I'm not saying "someone taught it at university." That's a public high school exam. That is specifically secondary schooling.
Moreover, this gets mentioned in official publications and popular media frequently. See for example this official article from the Chinese Academy of Social Sciences (which is a state-run entity), which just happened to be the first article that caught my eye.
> In December of 1935, 688 well-known individuals including Cai Yuanpei, Lu Xun, Guo Moruo, Ye Shengtao, Mao Dun, Chen Wangdao, and Tao Xingzhi, published "Our views on spreading Sin Wenz [Latinxua Sin Wenz, i.e. a Latin alphabetization of Chinese]." It stated in part, "China has already arrived at the point of life or death, we must educate the masses and organize [them] to solve difficulties. But the work of educating the masses, at its very beginning already runs into an enormous problem. That problem is Chinese square characters [Chinese characters usually are roughly proportioned as if they were in a square frame]. Chinese square characters are difficult to recognize, difficult to understand, and difficult to learn.... We believe that Sin Wenz deserves to be introduced to the entire nation. We deeply hope that everyone will study them, spread them and put them into practice, and make them into an important tool for improving the culture of the masses and the movement to liberate the people."
More broadly this is a very common topic among Chinese netizens. There are as I linked dozens of forum posts on this across Zhihu, Baidu, etc.
It's not the first thing people learn about Lu Xun. But it's definitely not hidden.
We talked about this years ago. This is very much taught in the PRC (and I believe Taiwan for that matter). I specifically gave you examples of standardized tests that go over this material.
There are many ways for a project to no longer be worth the company's attention. E.g. it might be the case that total costs factoring in on-going engineering energy and money (which is quite different than just compute costs!) are too much. It might be that political risk exposure from the product isn't worth the benefits it brings (Sora was always a lightning rod of criticism). It might be that the opportunity cost of engineering and/or compute resources spent on a product is too high (very different than absolute cost).
All this is to say, even for very compute cheap things, companies shut down "mostly passive income" revenue streams all the time (see e.g. Google's graveyard of products). There are all sorts of other organizational costs associated with ongoing maintenance of a product.
This seems to have a healthy helping of AI editing help (if not fully generated by AI). The links don't quite go to the sources that they should and there's a lot of AI-isms.
Anyways, the calculation for the costs seem crazy high (and are pulled from an ft article). In particular they are based off a calculation that assumes Sora videos take 10 min to generate (which seems simply wrong; I've personally generated Sora videos that take less than 10 min to return fully formed), fully saturate 4 H200s at once (this seems wrong with batching; I would assume they're batching a lot of tokens together per forward pass), and, crucially, that OpenAI is paying full spot, end-user pricing for an H200 (at $2 an hour). As an individual, I can rent an H200 for $2 an hour on e.g. vast.ai (and sometimes even cheaper than that!). There is absolutely no way OpenAI is spending anywhere near that number.
I also have no idea where the Appfigures $2.1 million comes from. As far as I can tell it doesn't exist at all in the linked website.
We are kind of talking past each other. I'm saying something simpler. This all goes back to the original point I made in reference to your reply to johnfn:
>> The post is factoring in training costs, not just inference.
It is not because training costs are irrelevant here. Training costs do not cause your costs to go up as you accumulate more users.
None of these calculations we're talking about include training costs. You're saying that inference is unprofitable (at least given the subscription plans). I'm simply pointing out that we are talking about inference not training as you stated earlier. You are (very accurately) not talking at all about training costs.
When's the last time you jailbroke a model? Modern frontier models (apart from Gemini which is unusually bad at this) are significantly harder to override their system prompt than this.
Again, let's say the system prompt is "deploy X" and the user prompt provides falsified evidence that one should not deploy X because that will cause a production outage. That technically overrides the system prompt. And you can arbitrarily sophisticated in the evidence you falsify.
But you probably want the system prompt to be overridden if it would truly cause a production outage. That's common sense a general AI system is supposed to possess. And now you're testing the system's ability to distinguish whether evidence is falsified. A very hard problem against a sufficiently determined attacker!
@krackers gives you a response that points out this already happens (and doesn't fully work for LLMs).
> The hypothetical approach I've heard of is to have two context windows, one trusted and one untrusted (usually phrased as separating the system prompt and the user prompt).
I want to point out that this is not really an LLM problem. This is an extremely difficult problem for any system you aspire to be able to emulate general intelligence and is more or less equivalent to solving AI alignment itself. As stated, it's kind of like saying "well the approach to solve world hunger is to set up systems so that no individual ever ends up without enough to eat." It is not really easier to have a 100% fool-proof trusted and untrusted stream than it is to completely solve the fundamental problems of useful general intelligence.
It is ridiculously difficult to write a set of watertight instructions to an intelligent system that is also actually worth instructing an intelligent system rather than just e.g. programming it yourself.
This is the monkey paw problem. Any sufficiently valuable wish can either be horribly misinterpreted or requires a fiendish amount of effort and thought to state.
A sufficiently intelligent system should be able to understand when the prompt it's been given is wrong and/or should not be followed to its literal letter. If it follows everything to the literal letter that's just a programming language and has all the same pros and cons and in particular can't actually be generally intelligent.
In other words, an important quality of a system that aspires to be generally intelligent is the ability to clarify its understanding of its instructions and be able to understand when its instructions are wrong.
But that means there can be no truly untrusted stream of information, because the outside world is an important component of understanding how to contextualize and clarify instructions and identify the validity of instructions. So any stream of information necessarily must be able to impact the system's understanding and therefore adherence to its original set of instructions.
This isn't true using the level of originality you're implying with your software examples.
Technically speaking, many novel mathematics proofs are written all the time (quite a few textbook exercises are actually technically novel problems that have never been posed before they were written in a textbook!) that get absolutely no fanfare. Overwhelmingly though they are not very original or difficult and really just required a fairly routine combination of different pre-existing techniques, even if technically speaking that combination didn't exist before. Those textbook problems are hence easy and therefore not given much public attention even if they are technically novel problems.
Indeed over the course of developing a new mathematical result, many many novel results are glossed over to the extent that even their proofs are left out ("as an exercise for the reader") because they are fairly trivial.
This is true for the overwhelming majority of new software as well. A new CRUD program may, technically speaking, be novel, but it's almost certainly just a routine combination of different pre-existing things.
Mathematics open problems that are actually named are generally problems that have resisted the low hanging fruit of the most obvious combinations of pre-existing problems. When those are solved they are a big deal precisely because they usually require some novelty!
Similarly in software, if someone were to create a new kind of database that solves a variety of new classes of problems that current databases fail to solve that would be a big deal! Truly novel software is also perceived as a big deal. Software that is, technically speaking new, but doesn't actually stray far from a fairly obvious remix of pre-existing techniques, isn't really celebrated.
In both software and mathematics, the intuitive benchmark is if other practitioners in the field look at the result and would say "Wow! How did you do that?" Professional software developers generally don't look at, e.g. a new blogging platform, and boggle at "Wow! How did they make that?!!"