I've implemented fluid mechanics using Claude (through Cursor) and it had no problem writing the logic and integrating it using my custom physics engine and custom renderer.

So no, I don't think your assessment is correct. LLMs shine when they get to implement something from scratch on a blank slate with clear API boundaries, whether it's a CRUD app or a physics simulation. Where I think they struggle the most is in big legacy codebases on tasks spanning multiple modules with lots of red herrings.

I'm a computational fluid dynamics (CFD) engineer. Like others have said, I'm pretty confident basic CFD algorithms are in the training data of many LLMs. I would say a bigger problem is the accuracy of the generated simulator. A LLM would not be able to generate good tests. You need both tests for the math ("verification") and tests for the physics ("validation"), and LLMs can't do either at the moment.

Gold-standard verification tests are constructed using the "method of manufactured solutions" (MMS), which can be largely automated with computer algebra software, but are still quite tedious. I know from experience. I don't believe LLMs can handle the algebraic manipulation here particularly well.

Worse, LLMs won't be able to produce actual experimental data to do the validation test with. You'll need to track down one or more experiments from the literature or do your own experiment. LLMs might in the future be able to point you to appropriate experiments in the literature, but they don't seem able to do that at present. I think LLMs might provide useful advice when a simulation ends up not matching the experimental data. LLMs seem to know a thing or two about turbulence modeling, though I would question their knowledge of the most recent advances.

(If you're only interested in fluid simulation for games or computer graphics then physical accuracy is not a priority. But you probably should still use MMS to make sure you've implemented the math correctly. MMS is an interesting technique that has no parallel in software testing in general. Abstractly, the idea is to make a minimal modification to the software so that you have an oracle, and the nature of the modification is such that the modified software passing the test implies that the unmodified software would also pass the test. This idea probably can be applied in other areas.)

You could go and try it out. You'd be surprised how capable the LLMs, specifically Claude are now at creative projects such as this

The "issue" with saying an LLM can't do this is that CFD simulations are not actually that niche. Many university courses ask their students to write these types of algorithms for their course project. All this knowledge is present freely on the internet (as is evident by the Youtube videos that the author mentioned), and as such can be learned by an LLM. The article is of course still very impressive.

undefined

Claude is pretty good at writing Python programs.

I mentor at a public high school that has a dozen metal lathes, a half dozen wood lathes. Plasma cutters. CNC everything. A giant planer. My favourite thing is that half of it is super modern and half of it is indestructible WWII surplus, which looks really cool mixed together.

I cannot imagine the concept of lathes being taken out of schools because of accidents. What a self destructive cultural attitude.

Access to a CNC mill and a CNC lathe should be a fundamental right.

Being CNC with an enclosure mitigates 99+% of all the safety issues and makes them 1000% more useful. With CNC, the most harmful injury is likely to be a broken bone because you're only in the envelope with servos active but not the spindle. With manual, the spindle or chuck can grab you and kill you or send shrapnel into you.

And I really wish we had a good, open source pick and place. Those still can't handle 0201 components and higher density BGAs even though those PCBs are really cheap nowadays.

I picked some random public high school in my area (twin cities, MN). Here's some classes they offer:

> Machine Technology 1 - Increase your knowledge and skills in the use of common hand tools, engine lathes, shapers, drill presses, milling machines, and grinders. In this class students will have hands-on experiences in elementary mechanical fundamentals and principles used in precision manufacturing industries.

> Welding 1 - Interested in becoming a welder? Learn how to use the art of welding in artistic and industrial settings. Develop basic skills related to safety, equipment set up, metal transfer, gas shielding and welding various types of metals. This class focuses on oxyacetylene and gas tungsten arc welding.

> Construction Trades 1 - Interested in a career in construction or learning valuable home maintenance skills? In this course you will develop basic carpentry, metals, electricity, and pipefitting skills along with safe usage of hand and power tools applicable to each skill set.

I think there was a period (90s-2000s?) where trade skills were mostly not taught in public schools, but they're thankfully working their way back in.

I used a lathe by myself at school when I was 12 or younger. But we live in different times I suppose

Id take it with a grain of salt.. I used to work with Dream It Do It in western New York. There were many schools with machine and carpentry shops.

I learned recently that Charlieplexing is kind of like the discrete version of the Radon Transform: https://en.wikipedia.org/wiki/Radon_transform

Awesome! I hope you'll post some write up or something on how you did it!

Ooo nice! Looks sweet

Yeap, Tim is some serious maker/tinkerer/creator. Besides the very good hardware things he makes, there are also some other software stuff: https://mitxela.com/other/

If you're into the custom PCB design, I highly recommend Carl Bugeja: https://youtube.com/@carlbugeja He has lots of super interesting projects using micromagnetics, motors, flappers, and the like.

If you liked the part where he machined the housing, check out Clickspring. I think it's fair to say he is the greatest machinist currently active on YouTube, both in videography and storytelling, and in actual skill. https://youtube.com/@clickspring

I of course don't know your specific circumstances, but nonetheless I'd like to recommend you the author's own take on their work described in a rant titled "Spare Time and Hard Work".

It's available here: https://mitxela.com/rants (unfortunately It's not possible to link directly to it so you'll need to scroll a little bit down to find it)

I personally like coming back to it from time to time, since it's inspiring in it's own way for me.

Don't be deflated, get started.

Seems like it’s sold out now except for a special one at £1,000. Did anyone catch what the original price was?

One sense of the word is “a first full-scale and usually functional form of a new type or design of a construction” (Webster)

I think a plastic housing might look really nice TBH.

£1k, on his website he has a link.

I'd be tempted to fill it with water with something dissolved in it to make it more conductive (and probably more viscous), and include LEDs whose brightness and/or color reflects the conductivity at their location. The rationale being, why should I do all that math when Nature will do it for me?

Water is a lot less luminescent than LEDs. I think the effect might be reduced if changed to a water based display.

At such a small scale it will not slosh around satisfyingly.

i was wondering whether this could be hardware accelerated.