Today, we're excited to share Cartesia Narrations, in public beta. Narrations is a creator tool for narrating long-form content using Sonic 2.0. You can use it to create audiobooks and podcasts, narrate your Substack posts, and more. Some highlights🧵

Lucas Beyer (bl16) the world if we could write F.pad(x, pad=((1,2), (2,4)), dims=(0,2))

getting your speculation accepted must feel good asf for the draft model

Excited to share our new paper, "One-Minute Video Generation with Test-Time Training (TTT)" in collaboration with NVIDIA. We augment a pre-trained Transformer with TTT-layers and finetune it to generate one-minute Tom and Jerry cartoons with strong temporal and spatial

Training-free acceleration of Diffusion Transformers with dynamic sparsity and cross-step attention/MLP deltas--collaboration with soham and Dan Fu! ⚡️ 3.7x faster video and 1.6x faster image generation while preserving quality! 🧵 Open-source code & CUDA kernels!

introducing chipmunk—a training-free algorithm making ai video generation 3.7x & image gen 1.6x faster! ⚡️ our kernels for column-sparse attention are 9.3x faster than FlashAttention-3 and column-sparse GEMM is 2.5x faster vs. cuBLAS a thread on the GPU kernel optimizations 🧵

But yuppy fishtanks are only useful if they actually catch the yuppies, and nobody wants to live at The Brady as long as it sits atop an open-air drug market. The city needs to address crime to solve its housing crisis. x.com/kane/status/18…

the only aligned ai is waymo

me rn

I converted one of my favorite talks I've given over the past year into a blog post. "On the Tradeoffs of SSMs and Transformers" (or: tokens are bullshit) In a few days, we'll release what I believe is the next major advance for architectures.

Tokenization is just a special case of "chunking" - building low-level data into high-level abstractions - which is in turn fundamental to intelligence. Our new architecture, which enables hierarchical *dynamic chunking*, is not only tokenizer-free, but simply scales better.

happy to announce that we've gotten rid of tokenizers! especially excited with what we've replaced them with: end-to-end trainable modules that not only learn to group characters into (sub)words, but can iterate to group words into phrases and further higher-order concepts see

Got downgraded on my flight to NYC this morning and everyone in 1st class was sitting sadly all by themselves, without a cat. I looked back into the rest of the plane - everyone was happily petting a cat. Was just a very interesting observation.

ai tractor goes brrrrr

Grok Aaron Rupar That was a better answer. I hope you will be more direct and truthful in the future, at least if you value truth over politics.

you can just not do things

friendly reminder that a price ceiling creates deadweight but also improves consumer surplus

Technological innovation can be a form of participation in the divine act of creation. It carries an ethical and spiritual weight, for every design choice expresses a vision of humanity. The Church therefore calls all builders of #AI to cultivate moral discernment as a

It’s about shots on goal. Interviews are high variance. Don’t feel like a bad interview is an indictment on your skills or worth. Keep grinding keep working. Do interesting things and people will find you.