what kind of ideas did you give?

1. Cut the cells instead of deform; .org branch has this cell-cutting thingy it uses for multiphase flows (Claude found that on its own) and it can be used to cut cells; problem: you get small cells, slivers, high-skewness etc. It's much more difficult to fix bad cells than to just 'not make' them

ah and then say it opensource but only for people that request it

I am looking for any actual academic response to this? All I've seen is news articles proclaiming something along the lines of "Navier-Stokes equations have been fully solved by ai" (lol). Their proof is way beyond my mathematical skills, so I'd be interested in seeing what an expert thinks.

I have watched a few videos about it, in principle the following: - this is for the full 3d incompressible navier stokes equations - human proofs seem to have focused on even further subsets of the equations - they applied a smooth differentiable (they call it a "nice" function) as a force term to the equations - the starting conditions were a fluid at rest, so the fluid was just u = 0 everywhere - after t greater than 1 the fluid formed a singularity, something that would require infinite energy - they way it does it is by forming a swirl that starts to stretch out perpendicular to the vortex core itself. It then elongates itself to infinity (they call it spaghettification), which is impossible

I mean its not surprising honestly, the equations describe continuums which arent a real thing. I feel like its not surprising that it blows up even when "nice" boundary conditions are applied. To describe those cases better you just need to move a layer down and start accounting for particles instead of viewing it as a "goop".

I will believe in agi only when it can one shot thin prism layers with shm

Nvm late to the joke lol

Yeah I tried doing something similar a few weeks ago and claude just kept reinventing shm

Terrance Tao has a bunch of tweets/mastodon posts about it

Those are a good read

For those of you interested: Theres going to be a talk tomorrow (inperson&zoom) https://cmsa.fas.harvard.edu/event/navierstokes/

Apparently openai has also solved another millenium problem 🫩

I saw the proof apparently it relies on a specific configuration of a vortex collapsing inward while stretching horizontally via a scaling law

The proof seems weird for a fluid, I would speculate that there is a gap somewhere, or it could be correct

But is it really solving if they had to tweak a force enough to blow up a velocity?

I think it will lead to smoke and mirrors

Or end up getting more investors

They proved the set up of the config, so either there is a circular assumption or a mistake somewhere or, it is correct which i would be surprised with

Yes this is most likely a very big marketing win for them

From memory, as long as the force term is smooth it's considered valid

I mean the whole point isnt to prove anything physical, just mathematical

Show that the set is bounded or not

And if not why does it still can result in physical values

Idk something like this, often done in weak form things

Yes, but media is blowing up up...Just like the velocity field

🫡

message has been deleted

Sadly that’s my final project advisor and the rest of my group mates use ai for anything rather than actual sources 😔

Hello everyone, I am Aurnob from Bangladesh. Mechanical Engineering student from BUET. I am trying to get into OpenFOAM for my thesis. Any suggestions are highly appreciated