Helium exists in great quantities in the 4 big planets, which unlike Earth have strong enough gravity to retain it.
Others have mentioned that some helium exists on the Moon, where it comes from the solar wind. The use of the helium 3 from there has been suggested for nuclear fusion, if the fusion of helium 3 became possible (it is much more difficult than the fusion of tritium with deuterium, which is the main approach attempted for now).
However, for fusion relatively small amounts could still be useful. For other uses the amount of lunar helium might not be enough, even when ignoring how expensive it would be to transport it from there.
If we have such advanced tech, and trip to big planets would seem economically feasible, I think we will be long beyond the point of desperately needing transporting helium to do such crazy trips.
The OpenBSD bug was more difficult for LLMs, because it is an integer overflow bug, while out-of-bounds accesses are more common bugs that are found by most models.
The OpenBSD bug was also found by GPT-OSS and by Kimi-K2:
The first condition for finding a bug is to actually audit the code where bugs exist.
When a human does that, this is a lot of work, which is often avoided. LLMs can simplify this, but you must use them for this purpose.
As the link above shows, using multiple older open weights models was enough to find all the bugs found by Mythos.
The improvement demonstrated by Mythos is that it could be used alone to find all those bugs, while with older models you had to run more of them to find everything, because each model would find only a part of the bugs.
Even so, I prefer using all those open weights models together, at a negligible additional cost, while Mythos is unavailable for non-privileged users and even when it will be available for more people it will be much more expensive than the alternatives.
Nevertheless, the distance between free models and Mythos is not so great as claimed by the Anthropic marketing, which of course is not surprising.
In general, this is expected to be also true for other applications, because no single model is equally good for everything, even the SOTA models, trying multiple models may be necessary for obtaining the best results, but with open weights models trying many of them may add negligible cost, especially if they are hosted locally.
The errors caused by radiation are extremely frequent and you definitely must guard against them, otherwise anything will fail immediately in space.
However that does not necessarily require hardware measures. It may be more efficient if instead of a slow antique CPU with hardware redundancy you use a fast modern CPU, even if it is more sensitive to radiation and even when it lacks hardware redundancy, but you do each computation several times, verifying that every time you get the same result, and if possible you use different algorithms or verification methods, to be able to detect some permanent errors.
This is what the Mars helicopter did. If it had used standard smartphone software, the helicopter would have failed instantly.
>>The errors caused by radiation are extremely frequent and you definitely must guard against them, otherwise anything will fail immediately in space.
I asked this in another thread but I will repeat it here - how come that their bog standard iPhones that they use for taking pictures with are still operating fine then? If like you said, "anything will fail immediately" - doesn't sound like that's the case? They have electronic watches with no radiation hardening, they have regular laptops with no radiation hardening.....I'm not saying that it's not a problem, but it definitely doesn't seem to be in the area of "immediately failing in space" if you don't have that.
As other posters have said, the personal devices of the astronauts are already used in spaces that are much better shielded against radiation than a typical satellite or the Mars helicopter.
Radiation shields add mass and volume, so it helps if the electronics is somewhat resistant to radiation, allowing for less efficient shields.
Even with the enhanced shielding, the personal devices experience errors from time to time, e.g. the photographs taken may have some wrong pixels and they sometimes have to reboot their laptops or smartphones, if weird behavior happens. Like others have said, these kinds of errors are not important, unlike in the computers that control the spacecraft, where errors are not acceptable, so those must use either hardware or software means to combat the effects of radiation errors.
Yes, but that wasn't the question. OP said anything that's not radiation hardened will fail immediately - to which I ask ok, what about all the stuff they brought up with them which doesn't seem to be instantly failing.
The radiation levels are much lower where humans live, otherwise they would not live for long. Without humans, thinner and lighter radiation shields are used, to reduce costs.
Theoretically, one could use the same electronic devices that are used on Earth, if one would add thick enough shields, but this is impractical, so one must make a compromise, by combining some less efficient shielding with devices more resistant to radiation.
The Mars helicopter had essentially no shielding, as it had to be extremely light to be able to fly in the Martian atmosphere.
Moreover, as explained in the parent article, the radiation levels are not constant. A great part of the radiation comes from the Sun, and that part fluctuates continuously (i.e. the so-called "space weather"). The electronic devices must be designed to withstand the peaks of solar radiation, even if the radiation levels are less than that much of the time.
The astronauts can shut down their personal devices, preventively, when there is a peak of solar radiation, or when they pass through the radiation belts.
Aren't the stakes a little different with an iPhone that you have for picture taking and entertainment vs the systems that manage your trajectory and life support?
The fact that a handful of devices hasn't failed is hardly proof that they can't. Hell, I've driven thousands of times and never actually NEEDED my seatbelt.
>>The fact that a handful of devices hasn't failed is hardly proof that they can't.
Again, that's not what I'm saying. I'm just challenging OP's assertion that any device with no radiation hardening will "immediately" fail, which clearly isn't the case with these devices. That's not me saying that radiation hardening isn't needed, quite the opposite.
(I have no expertise or knowledge of this area but...)
tl;dr: people need (heavy) radiation shielding, cpus et al can live without it
I'd imagine their bog standard iPhones and watches are generally in parts of the craft which have more radiation protection than others and, further, that it's probably only the parts where people are going to be that get that protection (due to weight savings, etc.) and if you can mitigate radiation problems by using a $30 CPU instead of a $2 CPU and save $100K of weight on radiation shielding on the CPU compartment, that's a no-brainer.
There are laptops with ECC RAM, but they are uncommon.
Otherwise, the effect of memory errors depends on the use case.
If the laptop or mini-PC is used as a router/firewall/Internet gateway, then memory errors are usually not important, because they would result in corrupted network packets that are likely to be detected at the endpoints of a network connection.
If the laptop or mini-PC is used as an e-mail server or a Web server, then a fraction of the memory errors may result in a stored file that becomes corrupted.
At the small amounts of memory typical for a laptop or mini-PC, unless the PC is many years old there should be no more than a few memory errors per year at most, and the majority of the errors might not result in file corruption, but sometimes they may cause weird behavior requiring a computer reboot.
Anecdotally, during the years I have seen on the Internet a non-negligible amount of big files, e.g. movies, which appear to have bit flips that are likely to have been caused by their hosting on servers without ECC memory. Fortunately, in movies a small number of bit flips will not cause severe quality degradation.
With more valuable data, one must use ECC memory to avoid such problems.
This is similar to the difference between using error-correcting codes and using erasure codes combined with error-detecting codes.
The latter choice is frequently simpler and more reliable for preventing data corruption. (An erasure code can be as simple as having multiple copies and using the first good copy.)
> “We can now make lots of swap gates with neutral atoms”, says Tilman Esslinger, “but of course we still need a few other ingredients to build a working quantum computer.”
No, it does not make the comparison unbiased, because other companies, like BYD, may sell in a month as much EV cars as Tesla sells in a quarter, but they are distributed over diverse models, so one Tesla model may indeed sell more than any other model, without this reflecting the EV market share.
E.g. for BYD the 2026 target is over 1.5 million exported EV cars, with more than that produced for the internal Chinese market. During 2025, BYD exported more than 1 million EV cars, besides the production for the Chinese market.
Others have mentioned that some helium exists on the Moon, where it comes from the solar wind. The use of the helium 3 from there has been suggested for nuclear fusion, if the fusion of helium 3 became possible (it is much more difficult than the fusion of tritium with deuterium, which is the main approach attempted for now).
However, for fusion relatively small amounts could still be useful. For other uses the amount of lunar helium might not be enough, even when ignoring how expensive it would be to transport it from there.
reply