Rota 🚪🧎‍♂️

Continually return to what you love. What is beautiful and good and true

any interesting rabbit hole leads to God

@Bonecondor What’s the collagen brand

Strongest Claude user

@aarnogau Delete this nephew

Ive been telling friends who want to do start ups. Easiest way to get revenue and raise a few M: 1. Find something the models are bad at that people care about 2. Make a bench for it 3. Shout on X that some lab is better than other lab. 4. Sell data for the bench to other lab

@mean_field_zane @ZachariahSchwab @andimgladofit Same mostly

@ZachariahSchwab @andimgladofit wdym Malliavin calc bench? My experience is that 5.6 nails it.

did every single dimwit pseud on this website get a memo ordering them to post about writing cuda kernels or something? yeah man im sure you really honed your hpc skills while writing race science regressions

Every year I discover new and unexpected ways my heart can hurt. There are more ways to love than I could ever have imagined, and more ways to suffer too. I can feel the urge to develop a shell of sorts, to expect less and feel less, to habitually approach people with skepticism or frustration. It feels like there’s only two options; harden like a nut, or open like a flower. Ive been stubborn and insensitive for most of my life, and **leaning into caring more** feels foolish. I can understand why most people go the other way. How humiliating to expose more and more of everything that matters to sheer air, to probing beaks, to businesslike wasps, to let yourself tremble and tremble again. But you have to, it’s a fucking criminal waste otherwise. We’ll all be dead and no one will remember, isn’t the point to experience as much as you can while you’re here? Why start building the coffin before you’re dead?

@Thomasdelvasto_ x.com/anne_alyst/sta…

"i think i feared rejection so deeply that i convinced myself being unknowable was safer than being seen, because if nobody could fully reach me, then nobody could fully abandon me either."

weird to think ML research once meant spending time learning about matplotlib Figure vs Axes

@LololL5887 AGI canceled

@pli_cachete Would love to hear your review regarding 5.6

Fable has triggered its filter on a geometric measure theory question 3 times in a row. I know you Ants can do better than this!

@LololL5887 Better

@pli_cachete How is GPT 5.6 in geometric measure theory?

never learned to transfer that insight. i was too blind to it, and then it was too obvious to me.

@tszzl @theojaffee @1a1n1d1y

@theojaffee @1a1n1d1y yeah but I say this now with some amount of resignation

the new openai models are so insanely good i don't even know where to begin explaining..

@doomslide @jconorgrogan No but I didn’t know how to write it more earnestly

@pli_cachete @jconorgrogan /s ?

We are absurdly lucky that x risk lobbyists keep fitting AI progress models to exponential curves when facts on the ground place us comfortably to the right of the sigmoid knee.

@Jabaluck Reminds me of the Gaussian correlation inequality. These sorts of attention-constrained problems are all over mathematical statistics

Current cutting-edge AI proofs are consistently of the form, "This was a hard and long-standing problem that turned out to be easier than expected if you search persistently."

@doomslide @jconorgrogan You’re so brave for saying it

@jconorgrogan All incremental. You can fit any task distribution you want provided you have the data + infrastructure to do so. Generalization remains so weak that you can mostly gauge progress from NVidia's quarterly earning reports. x.com/doomslide/stat…

The Well Just Dropped: 15 Terabytes of Pure Physics Gold Is Now Open Source The scientific AI world just got a massive upgrade.Polymathic AI, in collaboration with the Flatiron Institute and researchers from Princeton, Cambridge, NYU, Berkeley, Los Alamos, and more, has released The Well: a staggering 15TB collection of high-fidelity physics simulations. This isn’t toy data. These are real, expensive-to-run simulations across 16 different physical domains, including turbulent fluid dynamics, supernova explosions, magneto-hydrodynamic cosmic flows, acoustic scattering, and active biological matter. Until now, reproducing this level of data required weeks on national supercomputers and grant money most teams will never see. The Well changes everything. It’s purpose-built for training PDE surrogate models the AI systems that can replace slow, costly physics solvers with a single fast neural network forward pass. Everything is fully open source, easy to load with PyTorch, and ready to drop straight into your training pipeline. Researchers and builders can now train on world-class physics data without the insane compute barriers that used to stand in the way. This is more than just another dataset drop. It’s a serious accelerator for scientific machine learning.The future of physics-informed AI just got a whole lot more accessible.Get it here: polymathic-ai.org/the_well/