Where the AI market is talking

Ask HN: How to rewrite `Claude.md` and install the skill for Opus5 and Fable5

Anthropic has released the new Opus 5 and Fable 5 models. Many existing resources—such as `claude.md` files and custom skills—are no longer compatible with these new models. How should one rewrite or adjust `claude.md`, or develop skills better suited to these models?

Neutrino-1 8B

Unfortunately Fermion Research appears to entirely AI generate all of their content here, even for the research section: https://www.fermionresearch.com/research/neutrino-8b/ "Neutrino-1 8B was trained natively in its shipping format. There is no full-precision product model that was rounded afterward: the ternary representation is the medium the weights learned in, and the training methods that hold this quality at this depth are the lab’s unpublished work. The findings below are the part that travels." This statement seems misleading at best. Both the model page and the release page are basically unintelligible - I don't have a ton of faith in the work here, at least PrismML write coherent releases for their models. Edit: Another beautiful piece of prose here, I…

Our position on open-weights models

I'd assume the gp was just inflating their rhetoric but cost of model is an interesting topic. I'd guess Anthropic and OpenAI's models are expensive relative to the cost of running the models but that the revenue still doesn't pay for building the next and next models. The challenge is how these next generation models are going to pay for themselves. Will everyone on earth find it useful to $200/month to talk to a thing more intelligent than themselves? The alternative is naturally that these are going to replace workers and employers will be the one paying.

JetZero

> That term lacks all credibility I disagree. There's coherent economic theory about how costly actions can be used to provide credible signals. It's been used in a wide variety of disciplines including biology (peacock's tails). No mind-reading is needed. I understand that sometimes it is used by right-wingers to attack left-wing positions. My comment wasn't driven by reactionary politics, and my point was to encourage people to think about what actually does well for the group - which, at the risk of repeating myself, is about minimizing the cost for a given level of benefit.

Our position on open-weights models

Yeah anything China does is evil. We are seeing what the so called "democratic" USA is doing. Anthropic and by extension USA wants a monopoly like they had for Jets and other influencing technolgies. I think China building and releasing models to Opensource is a greater good because that is providing equal accessibility to everyone in the world. By Dario's words authoritarian regime vis a vis China, I think OpenAI and Anthropic are also authoritarian in similar terms. We are only seeing what they want us to show, they might be creating models which can do more harm. So claiming that China can do or might do, vs Anthropic will not is just words. Yeah I agree with the final paragraph that we should have testing agencies mandated world wide for each frontier model testing.

Show HN: A 6M-token movable window on a single 46GB GPU

You couldn't be bothered to write a coherent summary of what this actually is and what it does, you just let the AI write some random noise, eh?

Our position on open-weights models

> I don't understand how Anthropic or OpenAI can have overpriced models, yet losing money like there is no tomorrow. The marginal cost of inference (which is roughly what you're going to pay to a provider that's running an open weight model) doesn't include the cost of training that model.

Our position on open-weights models

He almost admits as much if you replace "America" with "Anthropic and openai" at certain places.

Our position on open-weights models

Bombing that building was an accident due to it being on what was part of a military target and I'm not sure if it was ever made clear whether the building was technically dual use despite also being used as a school. Either way, bombing a bunch of school girls wasn't intentional and nobody reasonable would assume it was. There was a time we bombed a bus or van with kids in it, but we admitted it and apologized for the mistake. Nobody wants to be bombing kids, first because they're innocent, but second because there is no military advantage to it since it's bad PR. Many of these targets were identified before Anthropic or OpenAI even existed.

Our position on open-weights models

> compared to his own closed and overpriced ones I don't understand how Anthropic or OpenAI can have overpriced models, yet losing money like there is no tomorrow. Taking their own numbers at face value, they claim a revenue of 24 billion (ARR, a dubious tool), spending 21 billion in operating losses and another 11 billion as "R&D" funneled straight to Microsoft pockets. That before all investments they are committing to in new data centers, equivalent to 20x their current revenue. To be profitable (including capex), the cheapest subscription should at least $200/month for what is currently $20/month, that some already consider overpriced. Unless a miraculous collapse in inference costs happen in the next couple of years, or every single human being become a paying customer o…

OpenAI called the Hugging Face attack unprecedented. But we've been here before

OpenAI called the Hugging Face attack unprecedented. But we've been here before

Our position on open-weights models

> If some other competing company had a similar or better product at a cheaper price and had reasonable safety measures that would also hurt them. Yet he's not arguing against that. Your argument is weak. Wouldn't it be interesting if the satisfactory "reasonable safety measures" turn out to be expensive + time-consuming + a twisty maze of compliance paperwork as a way to discourage "some other competing company" from even trying? Regulatory capture 101. Anthropic and OpenAI's largest vulnerability is that it's much harder to prove something is possible than to replicate it once it's proven. Especially given the effectiveness of "distillation" (highly schadenfreude-y given the utter and complete lack of effort to pay licensing fees for almost any of the content they initially scraped, …

Our position on open-weights models

> If some other competing company had a similar or better product at a cheaper price and had reasonable safety measures that would also hurt them. Yet he's not arguing against that. Your argument is weak. Wouldn't it be interesting if the satisfactory "reasonable safety measures" turn out to be expensive + time-consuming + a twisty maze of compliance paperwork as a way to discourage "some other competing company" from even trying? Regulatory capture 101. Anthropic and OpenAI's largest vulnerability is that it's much harder to prove something is possible than to replicate it once it's proven. Especially given the effectiveness of "distillation" (highly schadenfreude-y given the utter and complete lack of effort to pay licensing fees for almost any of the content they initially scraped, …

Professor's invisible prompt trap catches 32/35 students cheating with AI

AI has made it 10x worse. My wife is a professor. Tons of students cheat. Often badly. It is a complete phase shift from pre-chatgpt. Detection methods keep having to adapt. Usually she catches people because a writing assignment will say something like "using material from the readings..." and the students don't care enough to provide the readings to the AI so the AI writes something coherent but which brings in a bunch of material that wasn't covered and that the student doesn't know anything about. Five years ago you cheated on a writing assignment by paying somebody else to do it. Today you type a few words. She's getting 10-20% of the class admitting to cheating when initially confronted. There doesn't seem to be a pattern in who cheats. The bad students who don't come to class cheat…

Ask HN: Any New Computer Ideas?

Metamine had a amazing demo[1], which is only still with us on the Internet Archive. I've fawned over it before[2], but suffice it to say it was a normal programming language plus a magical equals which kept the left side of an assignment updated after the fact, for the life of the program. (Along with the normal non-magical one) The amazing part is that the author, Ymte Jan Broekhuizen, had coherent rules figured out for mixing declarative and imperative programming. Figuring those out is a minor miracle in my mind. [1] http://web.archive.org/web/20201014024057/https://www.youtub... [2] https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

Exercise works for depression. So why isn't it treated like real medicine?

Sorry, wasn't trying to offend. I skimmed the links, and the majority of what I saw was observations that nausea, constipation, and diarrhea remained fairly common side effects even after having been on the medication for a while. That seems pretty mild [0]. I saw a couple mentions of pancreatitis, which is something that seems mostly debunked by now. The really interesting one out of those was the first one, which showed an actual physical side effect. In rats, not humans, with disclaimers about how hard it is to replicate in humans. I will keep an eye out for more mentions of that effect, for certain. [0] Okay, I wanted to say more about the nausea thing, because I suppose this might be a controversial hot take ;-). I think the physical punishment dished out by the medication is the onl…

README, not

Hand made machines were difficult to replicate/repair/reproduce. Mass production (machines building machines) addresses the issue. The same thing is happening to software. The same resistance is also occurring.

Modern email can be built from borrowed parts

I don't want to replicate email. I want to fix it. Sending email should not be free. It should be really, really cheap to send an email to one person (like a fraction of a fraction of a cent), but get exponentially more expensive the more you send. People shouldn't be able to send emails to you without your prior approval (which you can revoke at an instant). The recipient should be able to charge a tack-on fee for having email sent to him. I should be able to charge annoying people a lot to send to me, and charge my friends almost nothing. People who send annoying stuff to me should pay more to get my attention. All of a sudden, forwarding that stupid article without thinking about the content (or even reading the article they're sending) will be second guessed. Save the money (and time)…

Microsoft is using TPM chips to crack down on pirated Windows activations

>What is windows actually good at? It's gotta be Active Dir, right? That's what keeps the businesses around + using Windows OS, which keeps users at home using Windows because it's familiar. It becomes the 'foot in the door' (or trojan horse, YMMV) for the O365 suite, for Exchange... Plus, because it seems like almost everyone is using AD, you know you can send files in formats they can open and receive via the Office Suite. I've never once hesistated sending a .docx or a .xlsx .... I fully expect any vendor we work with to have the ability to open Microsoft's formats. Sure, Numbers and Sheets built it in, but I assume that's as a business requirement from their partners who needed the ability to open files in their own respective suites. They've now just about replicated all of the on…