Co-Founder at https://xitter.cf/t.co/PTCss7iltZ. Prev. @LumaLabsAI (Realtime Video World Models), @GoogleAI (VideoPoet). Let's automate research!
Mountain View, CA
Joined March 2015
- Tweets582
- Following774
- Followers3.1K
- Likes2.9K
Dan Kondratyuk retweeted
We're introducing TasteBench, a new standard to measure Research Taste on real world ML tasks.
Existing benchmarks have limited datasets and eval data ultimately becomes part of training data. In TasteBench, we continuously generate new ML tasks on OOS data no model has trained on.
Dan Kondratyuk retweeted
Today we are launching EdotEnv, a Quant Neolab building toward RSI.
RSI needs a loop of increasingly difficult tasks, which markets naturally are: Trading well means markets become more efficient, this makes successful trading harder.
Reach out if you are interested!
Dan Kondratyuk retweeted
At our latest YC Paper Club, researchers and builders presented on multi-GPU kernels, intelligence per watt, heterogeneous inference, and more.
Thank you to our presenters:
0:00 – @FrancoisChauba1: The case for chip and kernel specialization
7:16 – @stuart_sul: Parallel Kittens - Systematic and Practical Simplification of Multi-GPU Al Kernels (arxiv.org/abs/2511.13940)
21:29 – @JonSaadFalcon: Intelligence per Watt - Measuring the Intelligence Efficiency of Local and Cloud AI (arxiv.org/abs/2511.07885)
31:05 – @MarkSaroufim: When Al Starts Writing Systems Code
47:04 – Misha Smelyanskiy: Why AI Inference Needs Heterogeneous Hardware
1:04:33 – @shacklettbp: Building a High-Throughput Game Engine that Runs ENTIRELY on the GPU (madrona-engine.github.io/sha…)
Dan Kondratyuk retweeted
LLMs can't trade & higher reasoning doesn't help.
we ran SOTA models for a 2y period. TL;DR: they suck & reasoning doesn't help.
- no model comes close to simple static baseline
- more reasoning ≠ better trading
- when losing money Sol trades less instead of better
details 👇
Dan Kondratyuk retweeted
Introducing Modify Video. Reimagine any video. Shoot it in post with director-grade control over style, character, and setting. Restyle expressive performances, swap entire worlds, or redesign the frame to your vision. Shoot once. Shape infinitely.
Dan Kondratyuk retweeted
Even I am astounded at these results! Using @LumaLabsAI Dream Machine Camera controls with Ray2 text-to-video. And I'll teach anyone interested in learning how I do it over the coming months (it's not hard) . Extends up to 30 seconds in this!
Dan Kondratyuk retweeted
3D Chalk Art – A New Perspective
Step into the illusion with #DreamMachine.
Where flat images become dimensional scenes.
Powered by #Ray2 Camera Motion Concepts.
Dan Kondratyuk retweeted
Camera Controls for @LumaLabsAI Ray2 AI video is out now and IT IS GLORIOUS!
First impressions: Just look at these results! Are you not entertained!?
Dan Kondratyuk retweeted
Art in Motion with #Ray2 Img-to-Vid.
Bring paintings to life and create moving masterpieces. Just drop an image into #DreamMachine and ask it to move. Your gallery in motion awaits.
Dan Kondratyuk retweeted
The image-to-video mode in Ray 2 is so badass that I just can’t stop using it. 🎥🔥 @LumaLabsAI
Every test delivers insane results… totally hooked! 🤯
Dan Kondratyuk retweeted
Introducing Upscale to 4K – enhance any #DreamMachine video up to 4K resolution with all-new stunning detail. Natively generate in 720p or 1080p, then upscale your creations to 4K for next-level clarity. Available now.
Dan Kondratyuk retweeted
The Ray2 API is here. Developers can now build and scale creative products with our most advanced video generation model yet. From cinematic realism to dynamic motion, the possibilities are limitless. Start building today lumalabs.ai/dream-machine/ap….
Dan Kondratyuk retweeted
We are hiring in multimodal language model!
Deepseek shows that there's no better time to join a startup with smart people and solid engineering.
At @LumaLabsAI we are exactly that. We are innovating to build the best multimodal intelligence. To go beyond language.
Job post👇
Dan Kondratyuk retweeted
Today’s #SOTD goes to @LumaLabsAI, @leoyakxi, @Jean_mazouni, and @tuhin for "Luma Dream Machine".
awwwards.com/sites/lumadream…
Your ultimate visual thought partner—turn ideas into reality. Ideate, visualize, and share your vision with the world.
Congratulations! 🏆
#technology #animation #interaction
Dan Kondratyuk retweeted
This is not real.
Completely text to video generated by @LumaLabsAI Ray2.
I used ~ 30 video clips about Marine Biology, sound effects by Epidemic Sound, edited with Davinci Resolve:
Dan Kondratyuk retweeted
These are not iPhones but iPhone knockoffs which were made by cavemen in prehistoric times. Here's proof:
Dan Kondratyuk retweeted
🥳Excited that Physics-IQ benchmark is out and you can use it to test your video model's physics understanding.
Website: physics-iq.github.io
Priyank gave a great overview in his post so instead I will share some of my favorite model outputs below!
Do generative video models learn physical principles from watching videos? Very excited to introduce the Physics-IQ benchmark, a challenging dataset of real-world videos designed to test physical understanding of video models.
Webpage: physics-iq.github.io/