nathanmcrae

Laser trackers (used for metrology) can also use spiral search to find retroreflectors. Although I believe newer models generate a flash of infrared and find them via bright spots in the resulting image from a camera. I imagine modern celestial navigation systems evolved similarly (minus the infrared flash, not very useful for stars).

Side question, but does anyone know why specifically 2^-32 is the target floor for attacker success (in footnote 3)? I found another mention of the 2^-32 target in [0], but I'm not even certain they're related.

[0] https://csrc.nist.gov/csrc/media/Events/2023/third-workshop-...

I have used this for throwing together some models for 3D printing. I've found it very intuitive, though I'm not sure how ergonomic it would be for complex assemblies.

I really like the space-key based command access and default shortcuts for all the commands.

Aside, but it's pretty neat that the author has been semi-regularly posting on their blog for over 20 years.

In particular, the Order-of-Magnitude section might be interesting: https://nathanmcrae.name/order-of-magnitude/

I would also toss differential evolution in the ring as a widely applicable and easy to implement technique.

https://en.wikipedia.org/wiki/Differential_evolution

How do you go about positively demonstrating such a system is air-gapped?

Aeronautical charts have incredible information density. https://skyvector.com/ for zoomable charts

No, 600 megatons (~E12 kg) is closer. There are E10 humans and we weigh E2 kg. More precisely: 6E1 kg * 8E9 = 5.6E11

I've run Nextcloud for years primarily for webdav file sync and calendar+contacts. It pretty much just works. That's not to say there isn't any jank, but it's mostly if you try the more obscure apps e.g. I've set up the phone track feature and it worked for a while, but I've had trouble with it recently. Probably fixable, but I just haven't spent the time.

I would say it is the core component of my self-(vps)-hosted setup and easily the most valuable.

Even though this isn't a fully-fledged industrial system, I like the thought they put into those aspects. Particularly, I want to make a homebrew PLC myself someday and the 24V IO board they made would be a nice starting point for that.

The reason you would trust and FPGA more than a SoC is because putting in backdoors in an FPGA is much harder. If an FPGA is backdoored to the point where any soft core (whose layout kind be randomized to some degree) is compromised, then that would likely be obvious from the kind of inspections Bunnie has written about: https://www.bunniestudios.com/blog/2023/infra-red-in-situ-ir...

What packages would you recommend for this kind of usage?

I think pen-and-paper one-time pads are an underestimated tool for private communication. Granted they are cumbersome and limited, but they provide almost perfect secrecy and bypass issues of compromised computers completely. And with some basic steganography (section h in the guide below has a good example), you can pretty easily hide when / who you're sending a message to. 'The Complete Guide to Secure Communications with the One Time Pad Cipher' is a really good resource: https://www.amrron.com/wp-content/uploads/2015/05/one_time_p...

Some others have mentioned SomaFM and the Apollo missions separately, but SomaFM has a station that combines ambient music and the Apollo mission audio: https://somafm.com/missioncontrol/

I'm currently doing this and personally I believe the dates and names approach is best (depending on your goals). The theory is that if you have a solid grasp of the coarse details like births/deaths/major battles then when you are reading about the more subtle ideas (like what factors caused the fall of the Roman empire) you will be able to couch those ideas in the concrete framework you've already built. Then those ideas will be able to stick better.

I've only been doing it for a year and change so we'll see how it goes, but I think it's a good approach.

I think people should seriously consider using something like passphrase2pgp [0] in addition to a hardware key like this. That way you can have a brain key (hopefully generated with diceware or equivalent) to tie together day-to-day keys like this to a more permanent identity. I'm honestly surprised that strategy is not more widespread.

[0] https://github.com/skeeto/passphrase2pgp

This is exactly the kind of thing I hope LLM chatbots will be genuinely useful for. Though, how often do you find it completely hallucinating endpoints / parameters etc. ?

I find a good way to link together ideas is to have them easily at hand i.e. memorized. Having a motivating problem is probably a better way to do this, but I've found that motivating problems which require concepts I would like to learn are not as readily available as I'd want.

For me it was lambda calculus. I remember in high-school going through the motions of beta-reduction but having no idea what it meant. Much later I think I finally saw an implementation of natural numbers and addition using bare lambda calculus and it clicked that you really can represent any computation with just variables, abstraction, and application.

True, but verifying proofs is usually easier (or at least more mechanical) than generating proofs. Personally I think this is the best possible use of AI: to have it generate systems that have verifiable properties in some formalized language. That way we can just come up with the properties we want satisfied (which, granted, is not generally trivial) and use the formalized language to verify that the properties are satisfied in the AI-generated code.

I don't know how all of the factors you bring up interact, but as far as the last point: the universe is indeed uniform and isotropic at a large enough scale (~300 million ly), called, nicely enough, the End of Greatness ( https://en.wikipedia.org/wiki/Observable_universe#End_of_Gre... ).

I find that using Everything search (https://www.voidtools.com/) makes me use the filesystem more like a tagging system. I still name files and directories meaningfully, but I don't worry about the hierarchy at all. Then when I want to get something, I just search (parts of) the terms I want and see the matching paths instantly.

Using the filesystem hierarchically now feels painfully slow and awkward.

The paper addresses your first point in section 7 'Model Rationale'. Essentially, anything that is subject to self-propagation is going to tend to expansionism at a large enough scale, superintelligence or not. The only thing that would prevent this is wide-scale coordination (e.g. a universal autocratic government). And if any species did accomplish this, they would just be one of the 'quiet' ones.

That's my take, they wrote it better though. That section is short, straightforward, and worth reading.

Yeah, that's an important point, but that seems to be an issue with the subject of study rather than this particular theory. At least their model has very concrete and explicit premises that could be used to reject it. Particularly the 'hard steps' model of evolution: I don't know how widely accepted that is (genuinely don't, it may be accepted) and, if it is in time rejected by the scientific community then that would discredit--if not falsify--the theory.