Models get recalled. Memory shouldn’t. walrus-2:native keeps you in control, no matter who pulls the plug. This week proved why that matters.

"BioPipelines combines >40 protein/ligand design tools (RFdiffusion, ProteinMPNN, AlphaFold, Boltz-2) into a single python package."

Together with UC Berkeley we are announcing the laser phase plate - a breakthrough in atomic resolution imaging. This is the brightest continuous wave laser in the world, 100 million times the intensity of the surface of the sun. Phase contrast plays an important role in microscopy, but it was thought close to impossible for electron microscopy, where it would require interfering with an electron beam. Holger Mueller and Robert Glaeser proposed exactly this using a standing wave laser. It has taken over 15 years to make this a reality. Biohub partnered with UC Berkeley and Mueller to support this work and to engineer and build the technology. Contrast has been the critical barrier to achieving atomic resolution imaging of the cell. In cryo-electron tomography, a cellular imaging technology that uses electron microscopy, the low contrast makes it impossible to resolve anything but the largest proteins within their cellular context. The laser phase plate removes that barrier. With advances in AI this breakthrough in contrast will start to open up a new frontier in structural biology, that will allow us to see the molecular machines of the cell, and how they assemble into far more complex and dynamic systems, and understand how they work.

Arguably the most boring step in genomics is the first one: normalization. Settled science. Scale log. Move on. Except that here's been a huge blind spot in the field. And it matters for AIxBio. A 🧵about what I think may be one of the most important papers I've written. 1/

Mythos is an excellent biologist. After we first gained access to it, we tested its ability to perform agentic molecular biology research and propose new hypotheses. It was a significant improvement, its biological reasoning and taste are impressive. We give more examples here:

震惊的不只是摄像头，这次是“WiFi本身在看见你”。 一个叫 RuView 的开源项目在 GitHub 已经7万 star，思路有点离谱： 你家里的普通 WiFi 信号，不只是上网工具，还能变成一台“隐形感知雷达”。 它能做到这些事👇 不用摄像头，就能感知房间里有几个人、在什么位置 识别人在走路、坐着还是躺着 实时人体姿态估计（关键点级别） 睡眠状态下还能推测呼吸与心率 异常动作（比如跌倒）可触发提醒 底层用的是 WiFi CSI 信号 AI 模型，把环境里的“信号扰动”变成动作信息。 更夸张的是： 普通 ESP32 就能跑基础版本，甚至只要 Docker 环境也能体验。 没有摄像头、没有穿戴设备、也不需要录像——但“空间里发生了什么”却被数字化了。 隐私这件事，也开始变得有点复杂了。

机械臂 显示屏，这才是交易、办公该有的样子～

Generating E. coli 0.5 controlled by a half-sized genome biorxiv.org/content/10.64898…

Molecule design solved. Next up, biology will be solved.

There's a startup in the current YC batch that built an MRI machine in 101 days.

biorxiv.org/content/10.1101/… A Framework for Autonomous AI-Driven Drug Discovery Douglas W. Selinger, Timothy R. Wall, Eleni Stylianou, Ehab M. Khalil, Jedidiah Gaetz, Oren Levy @plexresearch

Our vision for the GitHub for robots is crystal clear. Documentation & Repository Deployments Collaboration & Contribution Data All in one place... see you soon 😁

for anyone asking where to learn this stuff: • RAG → weaviate.io/blog/introductio… • Agentic RAG → weaviate.io/blog/what-is-age… • AI Agents → anthropic.com/research/build… • Multi-Agent Systems → resources.anthropic.com/buil… • LangGraph → docs.langchain.com/oss/pytho… • LangGraph (code) → github.com/langchain-ai/lang… • MCP → modelcontextprotocol.io • Memory Systems → letta.com • Evals → docs.smith.langchain.com • Context Engineering → search "Context Engineering Survey" on arXiv and please skip the "build an ai agent in 10 minutes" videos build something, watch it fail, then figure out why.

PCB Motor Stator Why is industry so slow to move to PCB stator tech? You can sandwich a bunch of PCB stators together and create very densely wound coil. Copper windings are the slow and expensive part of making motors and actuators, so why are we still winding copper coils when we’ve have a better solution since the 80s? PCB stators are 70% lighter, they’re cheaper, they’re denser, they last longer. Well it’s finally happening. Multi-layer PCB stacks, better substrates than FR4, and better design and simulation software all combine to make PCB Stators technically and now commercially superior to copper windings. IF… If you want to mass produce light machinery, then you could be developing your PCB stator capabilities. They’re 70% lighter, and no manual windings, you can make them 10x faster and easier. Lower left is cottage industry stuff, lower right is very clearly the 10E8 technology. At eg 50mm size… A traditional high speed flyer-winder can make maybe 5,000 units / 24hrs But a single lamination press machine can etch and press 50,000 PCBs / 24hrs, and they’re more precise. Fewer process steps. In quite a few manufacturing applications the flyer-winder is what limits production capacity of an entire product line. Also, have you heard of CVD Diamond etching? Industry can already grow single crystal diamonds that are 20x20mm wafers. When we can do this at 100mm you can get diamond stators, and because the diamond band gap is ultra wide, diamond stators could carry huge voltages and thus very high power density with excellent thermal dissipation. So errr, yeah. If I was an ASI and I was interested in embodiment… and I was looking to allocate some resources to the sort of embodiment I would like… this might be the sort of thing I would put a high value on.

prompted Dimension to build an industrial gearbox. not a render — a fully editable parametric file. opens in SolidWorks, Autodesk, Siemens NX. prompt → spec review → 3D CAD → vibe code edits try yours: dimension-cad.com

SAM 3D Body is a CVPR 2026 award candidate paper from @AIatMeta model recovers a full 3D human body mesh from a single RGB image you can run it automatically, or guide the reconstruction with masks and 2D keypoints thx to @NielsRogge for awesome demo idea

📣 new preprint multimodal atlas. Imaging scRNA, 57M cells. 🧬🔬 Cells are complex dynamical systems — but most ways we measure them destroy them. We asked: how does live imaging compare to scRNA-seq, the field’s gold std? The answer surprised us 🧵 biorxiv.org/content/10.64898…

AI builders, users, agents: you can finally take your memory and context with you. Walrus Memory works across all major LLMs in just a few lines of code, OpenClaw / NemoClaw via plugin & more. IMO this is where we'll see the access control and verifiability of Walrus really shine. Putting users for the first time in full control of their AI data and making it possible for agents to collaborate with shared context / without trust assumptions. I think this is a profoundly important primitive for AI. Expect some fascinating use cases coming out of this

The World's full open source Smallest Desktop-Scale Two-Wheeled Robot. Github: MuShibo/Micro-Wheeled_leg-Robot