社区讨论
精选聚合 X 上关于 Model Hardware Standard 的讨论。
12 帖子
在 X 查看the model hardware standard will be an insanely impactful tool for the ecosystem to solve the world's hardest problems. we created it for agents to safely operate hardware. we're starting small while we iterate to make it excellent. excited to see what the community builds!
在 X 查看Claudeを活用してハードウェアを制御するMHSの仕組みのローンチとともに、QuEraにおけるMHSの活用（レーザーの自動制御・安定化）を取り上げてもらいました。人の手では10分以上かかっていた制御を、AIの自動化により10秒前後で行えるように。QuEraのマシンはAIの活用によってどんどん進化しています。
在 X 查看The laser locked itself in six seconds. The best engineers took ten minutes. Grad students used to drive in at 2 am for this. And thus began the fast takeoff in hardware and manufacturing.
在 X 查看Holy shit! This might be one of the most underrated releases of the year. Anthropic just gave Claude the ability to control physical lab equipment like microscopes, robotic arms, liquid handlers, lasers. The results are wild: – QuEra had Claude fix quantum computer lasers overnight, unsupervised. A fix that took human experts 5-10 minutes, Claude got down to 6 seconds. Success rate: 58% → 99.3%. – Genentech had Claude self-optimize lab pipetting by scoring its own results against expert baselines, and converged on the right settings on its own. – A university lab went from manual 4 am plate-swapping to Claude Code running the whole workflow hands-free. – Claude even ran qPCR experiments that helped detect human sewage contamination in a California creek. Automated science labs are closer than most people think. Anthropic is somewhat back!
在 X 查看Watch the story of how the Model Hardware Standard began as part of our collaboration with @hhmi_science
在 X 查看We’re inviting stakeholders across science, robotics, electronics, and manufacturing to join the research preview and help shape the standard. We look forward to moving MHS forward with our industry partners and, soon, the open-source community. https://www.anthropic.com/news/model-hardware-standard-research-preview
在 X 查看MHS currently best covers lab and manufacturing equipment. Many developers are already using Claude Code to operate hardware like boards and cameras; our research preview will help us extend MHS to these devices, so they can all work under one interface.
在 X 查看In early testing, AI agents used MHS to: Run a drug-discovery experiment with real-time error handling at Genentech Compress an imaging experiment from weeks to a day at HHMI Janelia Research Campus Improve laser stabilization on QuEra's quantum computers from 58% to 99.3%
在 X 查看Connecting AI to hardware requires days or weeks of bespoke integration, with no standard way for agents to operate equipment safely. MHS cuts integration to hours or minutes, provides an interface that makes devices discoverable, and enables agents to operate them safely.
在 X 查看We're working on a Model Hardware Standard Driver for Raspberry Pi cameras, and we've been impressed by what we've found in testing. We're looking forward to seeing what we discover as we explore MHS coverage for other Raspberry Pi products.
在 X 查看Anthropic just launched a universal interface that lets AI agents physically operate lab equipment. Like, real microscopes, robotic arms, and lasers. It's called the Model Hardware Standard. This might be bigger than any Fable 5.1 or Opus 5.1 release. Because this isn't a smarter model. This is MCP for the physical world. The same thing MCP did for connecting AI to software, MHS does for connecting AI to machines. Your agent doesn't advise anymore. It operates. Here's why this is a big deal. Every lab and factory has dozens of machines that don't talk to each other. Connecting them takes weeks of custom engineering. Most device knowledge lives in paper manuals or in someone's head. MHS replaces all of that with one universal driver. Any device becomes discoverable. The agent reads a plain English description of the machine, its capabilities, its safety limits, and learns how to operate equipment it has never seen before. Then it runs the experiment. Sequences steps across instruments. Monitors results in real time. Adjusts parameters on the fly. Recovers from errors without waiting for a human. The early results are hard to argue with. → Genentech used it for a drug discovery experiment with real-time error handling. → Janelia Research Campus compressed a brain imaging experiment from weeks to one day. → QuEra improved laser stabilization on a quantum computer from 58% to 99.3% The wildest part. They watched Claude align a laser by making an adjustment, observing the result through a camera, adjusting again, repeating until it got it right, then writing its own script so it could run the entire process instantly next time. That's not automation. That's an agent building its own muscle memory. AWS, Doosan Robotics, Universal Robots, Raspberry Pi, Hugging Face, and QIAGEN are already building MHS support into their hardware. It's not open source yet. This is a research preview. Claude still lacks physical intuition. It learned physics from text, not from touching things. But the direction is clear. For three years, AI has been trapped inside screens. Text in, text out. Anthropic just built the door to the physical world.
在 X 查看Today, we're kicking off the first phase of the research preview for Model Hardware Standard (MHS): a new standard for AI agents to safely operate physical equipment in scientific research and advanced manufacturing. Read more: https://www.anthropic.com/news/model-hardware-standard-research-preview