Croissanthology
This post went through multiple AI fact-checkers! Clearly that was not sufficient. I think the comments are right that for questions involving a lot of physics I should’ve gone to human experts (am slowly drafting a version involving them rn). Unsure an AI fact-checker would solve much of anything; at least, it would not have prevented this post from being automatically promoted to frontpage.
(I didn’t pull the post down because I think the core of the post still stands, and I suppose it’s also interesting accidental jurisprudence for what LW norms around AI-heavy research should be, at least with early 2026 capabilities.)
My plan is to do more research by asking experts (which I did not do, and is a mistake I will not repeat) and then in a few weeks/months announce an edited version that I’m confident works. I don’t feel comfortable modifying the epistemic status until this is done and multiple people for whom it’s their job to think about these things announce “yep this looks good”.
(To be clear, I don’t think the core claim of the post is wrong; for instance, it matches up with the (few) reports done on grid vulnerability to solar storms that are out there (like Lloyd’s insurance).)
We’ll see how this one pans out.
I would certainly like to speak to him to correct this post. Do I do so by email preferably? He listed it in his about me page you linked.
I added a disclaimer.
Yeah I’ve eliminated the footnote’s sentence. It was way too shaky and didn’t even bring much to the post. The reason I quoted it at all is because I had read the claim in an Economist article I couldn’t find, and thought it’d be interesting to include it.
> This was exactly the response I was hoping you would not make. The problem is not the mere existence of a specific error, but what it says about the process as a whole.
Right… well I’m emailing researchers now. I hope to overhaul this post that way. I will definitely do so first next time.
> This does not give me any confidence in your results given that in the most trivially checkable places so far, complete ignorant amateurs here have already found serious misstatements.
This was a month ago and I’ve smoothed over errors since. A lot of bewilderment has since faded. I certainly hope that’s not due to LLMs talking authoritatively to me, but I have more reason to cast doubt on that now.
> and America represents less than half global R&D so that’s at least a 100% overestimate
The report this is from (F “225”) is about American losses to everyone specifically, not total R&D lost to the Chinese. So foreign IP isn’t relevant.
> And I don’t think it is front page material until some actual experts, not LLMs, sign off on it.
Yes this post did not go through anyone who actually works in utilities or a space weather expert; I think now it was a mistake not to run this through some first. Now that there’s an artifact of my research thus far, doing so is easier, so I’ll do that now and add an epistemic status marker at the beginning.
Do you have any object-level disagreements? I don’t actually find this meta talk to be useful. Also Twitter does allow you to see individual tweets if linked, if you’re still curious (it’s about the state of the internet’s ability to answer questions around solar storms).
multiple factual errors
These were: 1) multiple errors in visualization from the Gemini image 2) clarification on what matters more (in some respects) to the US grid being able to pull off what New Zealand did, i.e. grid operators or grid utilities 3) the 225 billion thing is defensible, I think, as I clarified. Given the amount of facts laid out I don’t think this is a particularly bad track record!
I did write about the epistemics of this project, back when it was in its beginnings: https://x.com/croissanthology/status/2014071957692531107?s=20
I tried finagling with the model to get these changed, but that only made things worse. I’ve edited the caption to reroute to this thread, as I think it’s a little valuable to leave the original as a state of current AI capabilities. Thanks for these corrections.
Yeah I nearly included a GIF from a Kurzgesagt video where the actual mechanism of the CME’s magnetic field “linking up” with our own is animated, and it shows the movement and shape correctly. But I didn’t want to go into depth concerning that specifically, and decided to omit it.
I think most people have already seen a representation of our magnetic field and can understand for the sake of that diagram that it’s an artist’s representation of the magnetic field for simplicity. Also the field isn’t even the interest of Gemini’s diagram, the electrojets are, and they’re shown correctly.
I went with 225 as the floor, all sources I could find pointed toward the same source (a study run by the Commission on the Theft of American Intellectual Property), which always said 225-600 billion dollars globally, citing China as the majority source of IP theft. It makes sense that China concentrates almost all IP theft: 1) the CCP has had a deliberate plan to do this since Deng Xiaoping, and Chinese courts will not react to corporate espionage on foreigners, so it’s not like you can sue them 2) China has a unique-to-it forced technology transfer clause in contracts Western companies make if they want a base in China 3) the literal military does cyber espionage on corporate targets.
At scale, it doesn’t make sense for any other country to try to do this, since often it’s genuinely much easier to just buy the thing you want instead of being unlawful and constraining your trade with the US. So I went with saying 225 billion and linking the pdf that explains a lot of this.
Yes! China is the largest user of HVDC, and it’s not even close. I hesitated whether to include China in the post, because what would happen to its grid is really interesting (though not really a model for others, which is why I didn’t add it in). Maybe I’ll do a Quick Take about this if people are curious. HVDC is pretty damn cool.
Yeah you’re right about this, my bad. I edited the post accordingly. Thanks!
So the bottleneck to an NZ style plan (which works for strong but not dramatic storms, like the 2024 Gannon) is data; there’s no centralized data collection mechanism like the kiwis set up, and we also have far fewer GIC sensors per capita. What New Zealand did is rack up that data over two decades to then have good rerouting simulation maps they were able to execute in 2024. The 3,000 utilities figure is relevant because installing GIC sensors and centralizing data is a matter for them, not the grid operators.
“Grounded” ^^
Anyway thanks!
When you least expect it, you find new chopped unc Tomás B content
You may not know this, but there’s significant literature built upon the couch-at-lighthaven disgusting-zevia-sipping conversation that initiated this:
Yes there’s the original @Tomás B. rubber souls post but also:
https://www.sexmoneyart.com/p/contra-rubber-souls-on-creating-a
https://tomasbjartur.bearblog.dev/contra-rubber-souls/
https://croissanthology.substack.com/p/contra-the-yapping-grovelers
The tl;dr of all this yapping seems to be: “rubber souls are kind of a stupid concept, likely to be net negative to human nature, also the way these ideas were framed was idiotic”.
Yet another instance of “anime harems rhetorically save the day”
I endorse this kind of self love. (See post.)
I’ve replied to this post here: https://open.substack.com/pub/croissanthology/p/my-therapist-tells-me-i-have-a-great?r=5ivlcb&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
(It’s messy short-form, far below the standards of LessWrong, hence the Substack link.)
In my experience, not too important, except that if you have extended thinking on it makes things a little slower.
Also, often you can make your system prompt shorter by just replacing phrases with shorter ones and seeing if it has the same effect, etc…
The way I do things is my first system prompt is always a long list of everything I think I want, and then after time I can prune it down and delete all my inevitable repeats.
The intended understanding of those words was “I knew almost nothing about this subject when I started researching this, and used LLMs to find most of the stuff I read to write this post.” If there’s an expert on solar flare grid reliability on LessWrong, I would very much appreciate they write a post about it—the reason I wrote this in the first place is because, given the breadth of the estimated disaster and the likelihood past studies have attributed to solar storms (e.g. Lloyd’s), it seemed weird to me that LW was lacking a post trying to get to the bottom of this. Note I don’t think I would’ve made fewer errors in this post had I relied on Google to find sources instead. I think my mistake was not emailing any experts at all, despite that being an available move.