Opus 5.5 is good in ways that surprises me every day! It's incredible how big of a jump in multimodal it is! Got Opus 5.5 to move my furniture into a house on Zillow 🏡 I gave it a video of my home and the Zillow link for a home. My prompts: 1. "Learn the furniture I have from this video, then make a 3D version of this house and furnish it with my furniture. Make it photorealistic" 2. "Now make me a walkthrough video and a 3D version I can interact with" 3. "Make a video of this entire session showing the steps you took" Watch till the end for the walkthrough of 3D scene 😅 (Disclaimer: I'm not actually buying this house, it's just a random zillow link!)

Since I am obsessed with painting these days I got Opus 5.5 to replicate Mona Lisa. This was my prompt: "can you search for the picture of Mona Lisa, then replicate it using mouse movement and brushes in JS paint using chrome?" It took 20 minutes for the drawing part using Cowork + Claude in Chrome. After it was done I asked: "Ok now can you make a video of exactly what you just did? Showing every mouse movement and changes?" Result is the video you see! :D More details in thread 🧵

Using Claude in Chrome product. Prompt: replicate this image (my profile picture) in jspaint, use brushes to replicate it as closely as possible. I want you to use brushes and clicking. Do not use javascript tool, but the final result has to be very close to this image. DO NOT CHEAT 1. I specified the do not use javascript tool because otherwise it uses javascript and it is faster but it seems like people cared more about clicking :D 2. The task took 45 minutes, but I don't think it's that much slower than competitors 😄 3. Prompt is important. (See thread)

The amount of work that goes into launches is way more than people expect. I'm constantly amazed at how everyone comes together and takes ownership over different parts of the releases just to be sure everything goes smoothly. This one was a long coming, user-driven update that we kept getting bugged about on daily basis. Really glad it's finally out!

55.7% -> 70.6% on OSWorld v2 (released just a few weeks ago) Genuinely would love to have a harder computer use evals. Happy to sponsor a computer use benchmark that Opus 5 gets below 50% on! If you have an eval that matches this criteria please reach out! Happy to chat.

We're hiring on the computer use team at @AnthropicAI 💻 Building inside Anthropic has been a crazy amazing intense but beautiful ride. We've been heads down building since our initial launch of CU in March, and the problem space keeps getting more and more exciting. We finally decided it is time to grow the team. Looking for product engineers and researchers. Specifically people who: are genuinely into computer use and full agentic workflows are high agency and comfortable figuring things out without a map are low ego and collaborative, that's Anthropic culture and we lean on each other a lot If that's you or someone you know, reach out. Happy to chat even if you're just exploring. (Link to apply below) P.S. If we don't have a mutual connection but you're interested, DM me with your ideal role and a product you've built that you're proud of. A video walkthrough is even better!