MostlyStable
MostlyStable on Hacker News
The internet seems like the
easy solution, not the
obvious solution. For years, we have used the internet for things that didn't need it because it was
easier and it seemed safe enough. I think that the coming months and years are going to see a lot of companies burned by that decision and scrambling to undo it.
The fact was that it was never safe, it was never secure, but that taking advantage of those problems was hard, and generally not monetarily worth it, meant we ignored the issues. That's no longer the case.
As for low quality software, the floor for the quality that internet connected software needs to have is now dramatically higher than the quality that non-connected software needs. I think that in many applications, it may no longer be economically worth it, or maybe even possible at all, to hit the necessary quality.
That's a good point. The coding specific question (which still uses the arena.ai leaderboard, just the coding version instead of text version), has Anthropic at 66% at Anthropic/OAI at just shy of 80%. That's probably more representative even if Arena.ai isn't my favorite benchmarking site.
Actually....looking in more depth I may have to upgrade to something stronger than "not my favorite benchmarking site". The first OAI model on there is GPT 5.4 at 31....that's not really credible.
https://polymarket.us/event/code1-2026-12-31
Maybe I misunderstood the point you were trying to make with your original comment, but you seemed to be suggesting that this piece of hardware could negate the risks (from the banks perspective), of allowing their apps to support non-secure devices. If your suggestion only applies to a very small subset of customers, then that doesn't allow them to put their app (which every customer can use), on the untrusted devices.
You are suggesting an entirely second piece of hardware when there are people in this thread claiming (IMO correctly) that many customers prefer an app over the browser because it's very mildly more convenient.
And there is a very large swath of banking customers who I wouldn't be surprised to learn that banks will earn less than $130 from over the course of a decade.
That's a bold claim. It was almost certainly true for models in 2023, when it was written. I think that the recent spate of mathematical discoveries at least significantly narrow the implications for even current models (one can argue that they are only re-mixing existing mathematical concepts to solve problems that humans initially generated, but in my opinion, at some point you are really narrowing what it means for an idea to be "novel"). And I would seriously consider taking odds against it for models within the next 3 years (since that's how long it's been since it was published).
We are basically already at the point where if what current frontier models can do isn't truly "novel", then the vast majority of humans have never had a novel idea in their lives. My three years from now guess is that if whatever it is that frontier models can do at that point isn't "novel", then no human ever will have had a novel idea.
Remixing existing work into something novel is ~ the definition of creativity.
Oh cool, where did they copy the solutions to hundreds of open math questions from?
I self host (I have a large overlap with the services in TFA), and love it. I will never go back and think that more people should self host (or at least...join small groups of people who designate one person to manage it for a small number of people).
And believing that the current AI economy might have a period of rapid devaluation in the near future is pretty reasonable. But, much like the original .com bubble, the idea that AI as a technology will not survive and thrive through such a pop, and become an ever more important part of our future, is a belief that I can't comprehend. Anthropic and OpenAI very well may go belly up. AI as a technology isn't going anywhere.
And as much as I'd like to believe that more people may self host more things in the future, self hosting will never, even if it was adopted by literally everyone, become as big a deal as AI, even if AI hit a wall tomorrow and didn't get any better than the current frontier models.
Depending on your location and power costs though, before the recent crises, it was probably generally worth it get newer hardware for the power efficiency gains. I have an old DDR3 dual-socket XEON E5-2650 system, and while it's more than sufficient for everything I run on it (and then some), its idle power consumption is pretty bad. If prices do ever come down, I'll probably try to replace it purely for that reason, even though I'm not sure I'll ever need more raw horsepower than it's got.
I had always wondered about the magrittr package name, but never enough to actually go find out. That's a cool origin.
The bigger HVAC win, in my opinion, if you have time-of-use billing, is to crank the temp when power is cheap (usually overnight), and, assuming your house is relatively well insulated, letting it ride through the rest of the day. This does lead to some funny things like the fact that I use heavier blankets in the summer than in the winter because I chill the house so far overnight, but for anyone whose climate is relatively moderate and with a relatively modern house, this can save quite a bit on power costs.
Admittedly, this will probably not work that well in a Minnesota winter, or an Arizona summer, but I think a surprising number of people could benefit.
Pangram says 90% Human. And this didn't come of as particularly AI to my eyes.
I'm guessing he's talking about a BIOS update causing bricked mainboards, to which they responded somewhat slowly. A descirption of the problem with Frameworks eventual response is here: [0] A charitable take would be that, because of the low number of affected customers, it took them a while to confirm that it was actually a problem for them to address. My personal issue with it was mostly that, as a company who is so much about repair and upgrades, that even their eventual response was to replace bricked mainboards, rather than offering the tools necessary to recover (even better would have been design decisions like having a backup bios chip to recover from botched upgrades).
As a framework owner, the situation is less than ideal, but given that they did eventually commit to replacing even out-of-warranty main boards, I very much wouldn't agree with calling it "debacle". Of course, maybe he's talking about some other issue that I'm unaware of.
[0] https://arstechnica.com/gadgets/2026/08/framework-responds-t...
Cool concept. As they mention, more work needs to be done on long term durability, although they saw no degradation in their (very short) testing period. To me, the humidity control seems more worthwhile/useful than the power generation, although that could be a useful solution for small IoT devices like door and window sensors that don't need much power.
I like it as a proof of concept, would love to see a more serious engineering effort to demonstrate what real world costs, durability, and aesthetics might look like (if significant portions of indoor walls are going to be covered in this stuff, it needs to be able to look decent).
What's the end goal of this "build ugly things and depress land values" strategy? If it's to drive out current residents and buy cheap, to what goal? If, once you have bought cheap, you stop building ugly things and start doing high value things.....then land values go back up and you have defeated your own purpose.
See: Nuclear power for an example of this.
The decision to not require seats is one of the best regulatory decisions I have ever heard of, and I wish this kind of thought process was more present in all kinds of places. The funny thing is that, I bet if it were used more broadly, given the absolutely enormous safety advantage that flying has over basically any other mode of transportation, there would be a non trivial number of other flight safety regulations that would get reconsidered.
They were already safer than coal...by several orders of magnitude. A Chernobyl level disaster
every single year would have killed fewer people than coal power. The level of safety you are describing has literally killed millions of people.
Commoditizing the existing reactors of 40-50 years ago would have lead to a dramatically better world than we currently live in, with cheaper power, essentially defeated climate change, better air quality, and fewer premature deaths associated with that air quality.
Ich glaube, dass Deutsche buchstabieren es "aktuelle", eigentlich.
(apologies for my broken learner German, I decided not to use a translator).
Wow, I expected it to be a DNA based split or something like that, but a genuinely new to science mammal is pretty cool!
I'm pretty wary of mass public surveillance. I'm much less concerned about companies monitoring the inside of their private property
As far as I can tell here the ghost portion is basically a non-sequiter. The waymo vehicle called the cops because it detected e passengers had a gun. The vehicle did not know or care about its status as a ghost gun or not. I'm sure that the Waymo terms of service prohibit carrying weapons into the vehicle. I'm pretty sure that if you were brandishing a realistic looking enough gun, an Uber driver would kick you out and call the cops as well.
I'm honestly curious about the "market keeps rejecting it" part. Yes, they sold far fewer units than the other models....but did they really not make money? I have made variations on this comment many many times. I completely buy that the opinions about phone form factor and features expressed on HN/reddit/online, and my own idiosyncratic preferences probably even more so, represent somewhat of a niche, and that your median phone consumer isn't interested in what we/I want. But the phone market seems like it's so large that it should be able to support a lot of niches! A tenth of a percent of the total market (~1,300 million in 2023 according to [0], 0.1% of which would be 1.3 million phones) should be totally viable as a business! I get that, maybe for Apple there is opportunity cost. Having more product lines adds overhead, and if the profit margin on those extra product lines isn't enough higher than the main lines to justify that overhead, they kill it. But why doesn't some smaller company (so yes, this mostly applies to Android) try to steal that fractional market share? I simply don't understand why it doesn't work to the point that now, when the market is so much larger than it was in the past, we have so much less hardware diversity.
[0] https://www.statista.com/statistics/263437/global-smartphone...
The problem with mostly-but-not-entirely autonomous is that humans will tend to treat it like entirely autonomous.
That's how several major AI advancements have happened. I have seen no evidence that that is the fastest way to make progress right now. I expect that, much like chess engines, it will not take too long before AI is significantly better than AI + human. But right now, my bet is that we are still safely within the window where an AI + human mathematician team is still better than AI alone (at least for the case where the human has learned how to work effectively with the partner....something that this event could possible be good for teaching).
The not needing an account is obviously not minor, but honestly, the UI is just better, by a lot. I made a twitter account a while ago using one of those throw away email address sites, so I have an account that I don't care about that I can use for twitter....but the experience just sucks. I don't know how any stands it, even if they don't mind needing an account.
Even if they don't figure out how to exfiltrate weights, someone will intentionally do this with an open model once open models are capable enough. If you ever think "no one would be so stupid as to...", you are wrong. Yes, someone absolutely would, and will.
Independent models living "in the wild" is approx. inevitable.
I think this has historically been true and is still approximately true now. But I think in the relatively near future (less than 5 years) there's a really good chance it won't be true anymore.
Once open models catch up to the current frontier in vulnerability exploitation, the cost to target people will go way down. In the past, the cost to hack a random individual person was generally high enough that if there wasn't some special reason to hack you in particular, it wasn't worth it. That may no longer be the case in the near future. The floor of what is acceptable security for the General Public probably needs to rise quite a bit over the next few years.
The argument about the seriousness of the topic seems to be to be arguing for taking into account human psychology and toning down rhetoric (note: not failing to make the claims, or dropping points, just changing tone) in order to make it more likely that you get listened to. To the earlier commenters point: it would be great if nothing other than correctness + importance determined likelihood of being listened to. But that's not the way the world works.