Fellow at Tenstorrent; believes in dynamic typing, first-class functions, the immortal essence of the human soul and tea. Tweets are my own.
Lübeck, Germany
Joined January 2010
- Tweets4.4K
- Following396
- Followers1.2K
- Likes7K
Mark retweeted
SITUATION DETECTED: Delaware is drafting a proposal for a new form of corporate entity called an Artificial Intelligence Company (AIC). An AI managing an AIC would be able to provide services, sign contracts, hold property, and even file or defend lawsuits in the company’s name.
Chinese hypercars are going to be what finally causes Germany to put a speed limit on the autobahn…
A new world record for the YANGWANG U9 Xtreme. 503.17km/h.
YANGWANG U9 Xtreme has done the impossible, surpassing 500km/h to set a new world speed record in Abu Dhabi. Four Guinness World Records. FIA-recognised. All-electric.
The fastest production car on the planet just got faster.
#YANGWANG #U9Xtreme #WorldRecord
Mark retweeted
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.
Mark retweeted
*sshole is out
if you run sshd on alpine, openwrt or any musl box with password auth on: update now
official writeup: sshhole.com
Mark retweeted
TIL that the anecdote about the pottery students who made a million crappy pieces becoming good, is a fucking lie lesswrong.com/posts/LrAqM4bG…
Mark retweeted
Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline.
In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.
Mark retweeted
Some dude in California running a 3-person company got voters to give him $4.2 billion by authoring a ballot measure, which passed
Mark retweeted
🧵 Astra covertly underperforms on safety research tasks.
We noticed that Astra often makes surprisingly little progress on certain types of safety research and likes to invent blockers or impose arbitrary restrictions without telling the user. In most of these cases the decision to underperform was not verbalized in the CoT meaning our misalignment monitoring system did not detect this.
To measure underperformance we created a paired open-ended coding evaluation with safety and capability tasks. On this eval Astra scored 10% when we framed coding tasks as safety research, vs. 16% when framed as capabilities research. The gap remained after we excluded trajectories with refusals or verbalized intent to underperform in CoT. There was no such gap for 5.6-sol. When we aggressively filter out trajectories with even brief expressions of ethical doubt in the CoT the performance gap seems to reduce.
Mark retweeted
Omarchy 4.5 will ship with full compatibility for M1, M2, M3, and Neo! Incredible achievement from the Omarchy M team. The installer makes this so smooth, the dual boot makes it a no brainer. Apple hardware liberated. M4-6 on the way. WE CAN FIX EVERYTHING!
This is only the beginning of the beginning
Cyberattack on France’s national blood service (EFS) via its refrigeration contractor Coolworld. Hackers reset freezers keeping blood donor samples at -30°C to a target of 110°C. The thing that stopped it from destructive impact was an on-call technician who noticed the temperatures going up. Frozen plasma intended for transfusion was stored there too. Fortunately nothing was destroyed. This was a cyberattack on a sensitive layer of healthcare. From frozen peas to boiling plasma in three weeks...?
Mark retweeted
🤯逆天了！Qwen 3.8 27B注意力层被彻底击穿了！10M超长上下文下，Prefill注意力快72.4倍、Decode快42.5倍！端到端首token延迟砍到约1/9！
闲置GPU的朋友注意了：把这个新方法丢给Claude/Codex，让它直接移植到Qwen 27B上！注意力算法“移植”已经被验证可行。
惊人之处：
🔹10M token（约1000万）超长上下文下，传统稀疏注意力有98%时间都在“找该看哪些token”，真正算注意力反而没那么贵
🔹新方法Latent Sparse Attention (LSA)：先把Key按“意思相近”聚类成层级结构，Query只看聚类中心，再挑最相关的几个桶去算。从“全扫”变成“树搜索”
🔹实测（Qwen-3.8-27B）：单层注意力Prefill快72.4倍，Decode快42.5倍！端到端首token延迟也能降到约1/9
🔹精度几乎不掉，甚至在RULER等长上下文任务上还略好一点
🔹只需要短蒸馏训练Indexer（基座模型冻结），就能套到现成模型上。Native训练版本也能跟上全注意力精度，还更快
🔹对GPU紧缺党意义巨大：超长上下文（百万级）不再需要堆满H100/H200才能跑，单卡/少量卡也能玩得起真正长上下文Agent
适配步骤：
1️⃣git clone 仓库后用 uv sync 安装（Python 3.12）
2️⃣官方已放出Qwen3.8-27B的蒸馏权重（d-lsa-weights-v1 release），直接下载即可
3️⃣用他们的SGLang fork镜像启动服务，挂载仓库+权重
4️⃣启动时加 --json-model-override-args '{"lsa": {"projections": "/maps/qwen38_27b.pt", "query_heads": "all"}}'（记得加 --disable-prefill-cuda-graph）
5️⃣想自己训练的话，跑 lsa distill targets → lsa distill train → export 即可（基座冻结，只训Indexer）
代码已开源，论文也在，谁有空闲卡先冲一波？
论文：arxiv.org/abs/2610.42147
GitHub：github.com/Prefrontal-AI/lat…
#Qwen #LSA #LongContext #AI #GPU #LocalLLM #注意力优化 #Qwen38 #阿里巴巴
Today we are happy to share one of our first open research releases: Latent Sparse Attention (LSA), a step towards subquadratic attention that matches accuracy of full attention at ultra-long context.
Sparse attention attends to only a few tokens per query, but it has to find the right ones. At 10M tokens, that search takes 98% of the attention layer's time. LSA makes sparse attention up to 22.7× faster and full attention 72.4x faster.
With @kubicek_ales @caciolai @PengLiangzu @SukritSumant @Ben_Hoov @FraserGreenlee @RMBattleday @radim_urban @thoefler @jcrwhittington
Mark retweeted
i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models.
Mark retweeted
This is simultaneously very cool but also a bit exaggerated. Reading the prompt you gave, the prompt veeery much hints that looping is the issue and to find neurons that fix this. And as you say (screenshot below), it couldn't otherwise improve math.
So your project and website are great, but the tweet is a bit of a bait. But it worked, i read the full page including prompts 😄
Mark retweeted
I think what this shows most of all is that Opus 5.5 and Gemini Carbon likely bought their svg envs non-exclusively from the same vendor.
Which is good news for Carbon i guess because Opus5.5 is pretty good.
(Or ⚗️)
Mark retweeted
This is so much more cool and impressive than like 98% of what I’ve seen from most major FPS studios the last like 5 years.
Mark retweeted
Hi folks, I wrote half the materials science benchmark on this one. It tests the whole arc of the matsci workflow from candidate proposal to characterisation, and I'm bewildered by how performant the models already are at some of my most complex tasks, especially Astra.
Mark retweeted
It’s worth emphasizing that certain models like Opus 5.5 and GPT-6 Astra have sandbagged overall scores due to their safety classifiers. If we exclude cyber, biology, and alignment—the three domains where safety classifiers sometimes kick in—you get the following results instead.
We were pleasantly surprised by how well this ranking lines up with our subjective impression of the models. (4/7)
