Ethan

[Ethan @eth_herron]

@pitdesi Why not Kennedy center?

[Ethan @eth_herron]

@sweatystartup When I tore my UCL I had to visit a PA before they'd give me an appt w/ Ortho. The PA wasn't going to give me the appt until I curled a soup can for a few days to see if it would get better.

[Ethan @eth_herron]

@drivelinekyle @DrivelineBB How do you get it to reliably use it immediately? No matter what I put in md file I feel like I have to badger it for awhile before it defaults to using it.

[Ethan @eth_herron]

@afshawnl Only rep you can naively use with pretrained image/video models. Eventually we’ll optimize sdf designs in latent space of physics foundation models. It’s going to be incredible

[Ethan @eth_herron]

final results from yesterdays brawl with docker:
>150GB of containers on disk

[Ethan @eth_herron]

@pfau I feel like its pretty easy to argue that's what this is just with a high fidelity decoder back to pixel space. If there is a long-term planning model hooked up to this, it'll probably only access those latents, not the frames. at least in the short term

[Ethan @eth_herron]

@cgarciae88 Adding “the best way to help me right now is to be ruthlessly objective” solves this problem nearly every time and usually outputs something pretty insightful.

[Ethan @eth_herron]

@willdepue imo they’re way cooler once you know how they work.

https://t.co/fCIedj0Aub NeRF: Neural Radiance Fields

[Luke Darlow @LearningLukeD]

Learning where to look, learning what to see.

Work on Continuous Thought Machines continues to be exciting! As a teaser, take a look at our approach to "Foveated Active Vision". Check it out 👇

(details in the thread)

https://t.co/vUttjVWP8a

[Ethan @eth_herron]

NOTEBOOKS

[Cursor @cursor_ai]

Cursor 1.0 is out now!

Cursor can now review your code, remember its mistakes, and work on dozens of tasks in the background.

[Ethan @eth_herron]

@rabois @paulg yeah and correcting them with the market not blindly slapping tariffs on everyone

[Ethan @eth_herron]

@rabois @paulg So Buffett… who calls them an act of war

[Minh Nhat Nguyen @menhguin]

@JeffDean @jeremyphoward deepmind guy discovering normal consumer GPUs

[Ethan @eth_herron]

I’ve been using cursor way too much. I’ve cmd+enter 3 times now in a ppt

[Ethan @eth_herron]

@kalomaze As for undergrads yes it’s a problem. But I think that will be less so in the future once previous gen chips become cheaper and schools have time to further build out infra

[Ethan @eth_herron]

@kalomaze Well there’s other clusters out there usually funded by nsf that are far better. TACCs vista has gh200s and Delta at uiuc has a100s. Grad students can apply for grants to get compute. I’ve also used national lab clusters before but is slightly more difficult for int. students

[Simo Ryu @cloneofsimo]

@kalomaze That wont change per-request latency (unless you have extremely large speculative decoding going on obviously)

As Warren Buffet said, "9 pregnant woman won't make a baby in a month, unless she is a diffusion model"

[Ethan @eth_herron]

@cloneofsimo Feels similar to variational rf

[Ethan @eth_herron]

@shah_bu_land Did you naively train it with l2 or did you use their 3 phase scheme?

[Ethan @eth_herron]

@afshawnl Solid and physical modeling (SPM) is fun. Great community and covered a wide variety- design and manufacturing