Lukas Wolf

We invented a recursive megakernel generator called Hyperkernel, the most complete solution to megakernels and GPU efficiency. Hyperkernel-generated megakernels achieve state of the art performance on 11 different domains: inference (DeepSeek, Kimi and GLM), pretraining (Qwen), hardware fault recovery, robotics (π₀-FAST, ACT) diffusion and real time world models (Matrix-Game), full self driving (OpenPilot), voice models (Moshi), genomics (Evo 2) and protein folding (ESMFold). All of this in a single kernel launch. For context, Hyperkernel generated fully optimized megakernels for robotics, genomics and voice in under 50 minutes, and doubled the efficiency for >5 models generally using only about 60-100 lines of code per megakernel. This speaks to the unparalleled standard for code quality in the automated generations. Our first release today open sources the megakernels themselves. Over the next couple weeks, we will release the entire Hyperkernel stack, along with the complete serving stack for Kimi and GLM (20-30% inference efficiency gains across all concurrencies). Out of the box GPU machinery performs terribly for unorthodox research. We built Hyperkernel only 20 days ago for internal use; but given its multi-modality, we have decided to open source it. Creative research should not be bottlenecked by the lack of dedicated infra teams. Hyperkernels offer a one stop shop infra stack for all research. We would love for the open source community to add to it :) GitHub repo: github.com/Hyperstitioncc/Hy… NOTE: We primarily use NVIDIA B300s, but our kernel generation extends to other hardware (B200s, AMD, TPUs). 1/9 We absolutely crush NVIDIA and achieve the best low latency inference for DeepSeek v4.1 Flash. Compared with NVIDIA’s published vLLM results, Hyperkernel delivers: • 36.7% higher per-user generation speed at concurrency 1. • 28.2% higher per-user speed while serving two users, versus NVIDIA serving one. • 15% higher per user speed while serving four users, versus NVIDIA serving one. • 90.6% higher per-user generation speed at concurrency 4, compared with the highest published C4 AgentX interactivity for TP4 B300s. • 28.5% higher per-user generation speed at concurrency 8, compared with NVIDIA's published result. • 36% faster median per-request decode speed.

🚨 CRITICAL: GitLab flaw (CVSS 10.0) affects 12 years of releases. CVE-2026-110264 could allow unauthenticated remote code execution on self-managed servers. Read - thehackernews.com/2026/10/cr…

We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI. Post-training results coming soon -- stay tuned! Lots of details in our technical blog post. microsoft.ai/news/frontier-p…

launched a model accidentally served trillions of tokens a day 29.4% of the fortune 500 showed up raised a really big series A from @a16z it's been 3 weeks we would like to sleep now

We’re on a mission to build a billion AI models, not trained by us, but by you. Introducing @bryel_labs we help your company build its in-house AI lab and give you the software to make model development and continual improvement simpler. We are backed by @ycombinator and have trained models at Apple, Framer, and F500s. We’ve enabled companies to train: • Classification models that are 20x faster • Search models that cost 5x less. • Long horizon agent models that match the frontier at 5-8x lower cost. Are you interested in partnering with us? Reach out. bryel.ai/

SFT is not dead! 🥳 We found a way to make SFT rival current prevailing posttraining methods, often generalizing better and forgetting less than RL and OPSD. 🤯 Following our prior work on reasoning with sampling, we now introduce sampling to the posttraining stack. 1/n