It definitely was a real skill; there was a time when you could get impressive results out of frontier LLMs if you knew the most advanced tricks for prompting them, but not if you just told them "hey solve this problem for me".
However, it's not a sustainable skill, because the labs care a lot about making "hey solve this problem for me" work well, and so put out models that are better and better at working with unsophisticated prompts over time.
I recommend reading https://keymaterial.net/2025/11/27/ml-kem-mythbusting/, which explains (1) why it's not possible for ML-KEM (the thing most people are using for post-quantum encryption) to be backdoored the way Dual_EC_DRBG was, and (2) some of the reasons some people don't like hybrids, not all of which are good but none of which should serve to undermine confidence in ML-KEM's security.
I'm curious, what are you hoping to convey by reminding people that LLMs are next-token predictors? They are, of course, but most people without an AI background won't fully understand what that means, so I assume you're using it at least partly as a proxy for something else.
I think understanding how this stuff works is really important. For technical people it gives them a useful starting point for understanding it all. For less technical people it's crucial to help them understand that it's not some weird new magical science-fiction AI - it's still computer programs that turn text into numbers and do stuff with the numbers and turn those back into text.
It's harder to believe something is conscious or threatening to achieve word domination once you understand that it's a machine that statistically figures out which word should come next.
While you’re active in this thread, I just want to say thank you for all your writing, you’re such a reliable source of sanity in that crazy new world :)
I don't think the next-token-predictor thing should increase anyone's confidence that LLMs aren't conscious or can't escape the control of their operators. A very closely analogous argument would "prove" that humans aren't conscious or can't do [insert task here] either. (No, I'm not saying that any of this is true of today's LLMs, I'm saying this particular argument doesn't work.)
You can say that for any argument regarding consciousness, because we don’t have an actual, all encompassing definition of what consciousness is. In general I don’t think comparison with humans makes much sense, we should be able to discuss LLMs without always falling back to “but what about humans” (sorry for the caricature)
Shouldn't that imply that agnosticism is the proper view, rather than asserting that something is impossible on a next-token-predictor architecture?
(Note: I don't actually think the consciousness question is the most important one in the near term. Where I think this line of reasoning gets really dangerous is when people use it to assert that LLMs can't or won't engage in certain behaviors no matter much they advance; this doesn't have anything to do with consciousness.)
Apple absolutely does not provide Apple Card in-house, and in particular is not loaning you the money when you use it. Among many other reasons, for that to be legal would require Apple to be a bank, which they're not and they definitely don't want to be, because being a bank is a huge pain, and in particular means signing up for an extremely heavyweight regulatory compliance regime.
Apple Card is the result of a co-branding deal between Apple and Goldman Sachs (which is currently in the process of exiting the deal and being replaced in that capacity by JPMorgan Chase). Apple provides the branding and some minor fintech features; Goldman (or in the future Chase) issues the card and is on the hook if people don't pay their credit card bill.
Presumably for this particular program using Klarna was cheaper because Klarna already specializes in precisely this kind of transaction.
> because Klarna already specializes in precisely this kind of transaction
I think Apple gets sweetheart deals so their partners often lose money. Klarna likely sees this as a marketing coup to have their predatory sussness promoted by the Apple halo brand.
If there was profit to be had, I'm sure Apple could buy a small bank and spin it off as a subsidiary with little drama, it's just not the best of their options.
That last sentence is not true; we have gotten much closer to building a quantum computer that can run Shor's algorithm. Organizations like Google and Cloudflare have declared a 2029 deadline to completely stop depending on the security of pre-quantum algorithms; hitting that deadline is going to cost a lot of engineering resources, but they're paying that cost because they think there's too great a chance that nation-state adversaries will have scalable quantum computers by then. See https://words.filippo.io/crqc-timeline/ and the various posts linked therein, including from the aforementioned companies.
I don't trust promises from the ones who stand to gain from people believing them before they are delivered. Yes they claim it's jsut a couple of years out... we'll see then
Are you saying we shouldn't trust cryptographers, because there'll be more demand for their services if we need to urgently migrate to post-quantum algorithms? That seems a very radically-skeptic perspective that makes it hard to know anything, and I doubt any serious engineer applies it in the general case. Or were you talking about someone else?
In any rate, nobody (or at least none of the people I've heard from) is claiming we'll definitely have CRQCs by 2029. They're saying there's a real chance that we will, and that that means now's the time to pull the trigger on post-quantum migrations; if we wait for certainty, it'll be too late. Quoting Valsorda from the post I linked above (which I strongly recommend reading in full):
> If you are thinking “well, this could be bad, or it could be nothing!” I need you to recognize how immediately dispositive that is. The bet is not “are you 100% sure a CRQC will exist in 2030?”, the bet is “are you 100% sure a CRQC will NOT exist in 2030?” I simply don’t see how a non-expert can look at what the experts are saying, and decide “I know better, there is in fact < 1% chance.” Remember that you are betting with your users’ lives.
> Put another way, even if the most likely outcome was no CRQC in our lifetimes, that would be completely irrelevant, because our users don’t want just better-than-even odds of being secure.
She also notes, as a side remark in a different post: "I would say that my trust in lattice cryptography is pretty much equal to my trust in elliptic curves, and quite a bit higher than my trust in RSA." (https://keymaterial.net/2025/11/27/ml-kem-mythbusting/)
> When you download Python from http://python.org (on Linux or macOS), what you're actually downloading is an installer that builds Python from source on your machine.
> The net effect is that on Linux and macOS, you can't "download a Python binary" from... anywhere. Other than the python-build-standalone project.
Are you sure this is right about macOS? I just had a look inside the macOS installer from python.org and there are definitely some binaries in there.
It's definitely wrong for macOS. The python.org macOS download has been an installer package that installs Python to /Applications as long as I can remember.
(That said, I've rarely used it, preferring either the macOS/Xcode-provided Python or these self-contained Pythons which I use via uv.)
> (The notable exception here is that Linux distros _can_ support, e.g., `apt install python3` and ship a binary, because IIUC they can rely on standardized system paths.)
Of course, there are lots of use cases where that doesn't suffice, such as, for example, if you care what Python version you're using.
i think that sentence is taken out of context a bit, the problem with those pkg binaries is that they're not portable—which the post describes
EDIT: actually no, I think charlie is wrong on that point, he said the installers compile python but that only happens with python-build-based installers like pyenv. still, the fact those binaries lack portability is the real problem.
reply