working on inference. broadly interested in AI. sometimes writing blogs.
Joined October 2023
- Tweets784
- Following258
- Followers1K
- Likes6.7K
we certainly don’t live in a bubble :))
congrats to the team for great work!
we open sourced the @flappyairplanes megakernel.
they raised $1B on the back of a demo video: one GPU kernel that trains a model, serves it, and plays Doom at the same time
me and @Bobliuuu reproduced it from just the video frames and open sourced it; see this thread for details 📷
code: github.com/Bobliuuu/open-fla…
video below
Simon V retweeted
We’re introducing Hydrogen-1.0, our first open-weight model for scientific research. We post-trained GLM-5.3 (@Zai_org) on our internal scientific data, lab records, and discovery trajectories.
It scores 38.3% on FrontierScience-Research vs GPT-6-Astra’s 33.3% and scores 93.0% on GPQA-Diamonds.
We believe science should be open, collaborative, and democratized, rather than proprietary. Only open weights let researchers inspect, adapt and build on the unprecedented speed of AI growth. There are many recent models named by chemical elements, but we believe Hydrogen-1.0 is the first aiming towards and open scientific superintelligence for discovery. 🧵
it seems confusing to me that many AGI pilled people are very much against the welfare state and taxes. who will provide means of living to the masses after there are no jobs left for humans and no taxes are being paid?
recently playing around with the deepseek harness in order to avoid depending too much on claude code and the experience is pretty nice :)
i don’t understand why people try to assess if open source model provided via API is quantized via benchmarking. there are far more accurate ways to do that like checking distribution
Simon V retweeted
This piece of art is meant to show Jesus overcoming war & nuclear holocaust.
We can argue about artistic taste - it's not my favorite style - but it clearly shows Jesus rising over the chaos of world, which is a strong Christian message.
i think much more interesting than competition for optimisation of isolated kernels would be one where participants have to surface serious bugs for an inference engine. much less automatable currently and thus more real world value.
Simon V retweeted
Decisions affecting human life, freedom, and the use of force must always remain subject to effective, informed, and responsible human oversight. For this reason, artificial intelligence must be “disarmed”, and I encourage the States meeting in Geneva in November to take concrete steps toward banning lethal autonomous weapons.
Simon V retweeted
The future of #ArtificialIntelligence is not pre-determined. As a technology developed by human beings, its future depends on the choices we make. We therefore have the responsibility to guide, change and, when necessary, reject its applications. A reliable system that serves the human person is of greater value than one that is exceptionally powerful but whose consequences cannot be adequately foreseen or controlled.
benchmarks are slop and everyone who is using them extensively in his or her day to day work knows that.
Simon V retweeted
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity.
💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis.
🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others.
Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it.
Read more from our team and customers 👇:
Simon V retweeted
I improved NanoGPT Speedrun from 39.9s → 6.7s (-83.2%), more than every record since 2025 combined (-81.4%)!
The web is full of rare, repeated content that the LM can’t memorize in time. I solve this with retrieval-augmented n-gram hints🧵
i suggest having children as a constant and AGI proof source of joy to programmers coping with the fact that they don’t code by hand anymore
this is so cool and makes me finally read already-clean.goodfire.com/
Simon V retweeted
vLLM matches DeepInfer, DeepSeek's own engine for DeepSeek-V4.1-Flash, on DeepSeek's decode configuration across batch sizes 8 to 256, and does it at DEP32 against DeepInfer’s EP64DP64.
DeepInfer was originally built on vLLM. By integrating DeepSeek's open-source kernels and extensively optimizing prefill and decode, vLLM now matches DeepInfer's performance curves.
