🚨 Anthropic has a new CTO: Rahul Patil, the former CTO of Stripe. Patil started at the company earlier this week, taking over from co-founder Sam McCandlish, who will move to a new role as Chief Architect. Read more in @TechCrunch from @russellbrandom

Introducing Claude Sonnet 4.5—the best coding model in the world. It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.

Claude Code Office Hours nitter.app/i/broadcasts/1vOxwdBqR…

Claude can now use Skills. Skills are packaged instructions that teach Claude your way of working.

the rumor is openai drops “agent builder” tomorrow and wow, if that's true thats a BIG DEAL for the 12 months, people have been stitching together tools like n8n, zapier, make, vapi, and claude workflows to simulate autonomy. it worked but it was duct tape. now IMAGINE that entire stack, native to openai, with one-click access to MCP, chatkit widgets, and every model they’ve trained (no API chaos. no patchwork. one smooth canvas) this is what happens when ai moves from tool to infrastructure. before: you needed 10 tabs, 5 plugins, and a weekend to build an agent. after: you’ll drag a few blocks, add logic, hit “publish,” and deploy a production-ready workflow. what app store did for software, agent builder COULD do for intelligence. it’s the beginning of the “no-code ai economy,” where building an autonomous agent is as simple as building a notion template. developers get leverage. non-technical founders get superpowers. businesses get workflows that run 24/7 without ops teams. openai might launch the app store for intelligence tomorrow. the DOWNSTREAM effects: - zapier and n8n lose their monopoly on automation - claude and perplexity become upstream research assistants for agent networks - indie agents replace indie apps – data becomes the new code tomorrow's dev day should be INTERESTING.

Making an Email Agent using the Claude Code SDK If I wasn’t at Anthropic, I would be making agents using the Claude Code SDK. But doing > talking. So I’m building in public and open sourcing a local email agent. This is part one on agentic search.

I've been using Claude Code completely wrong. I gave it a custom skill and Browser CLI tools and letting it do work for me. It can open pages, click buttons, fill in forms all from your authenticated browser. Just published it to the marketplace, install it in 2 commands.

New on the Anthropic Engineering blog: tips on how to build more efficient agents that handle more tools while using fewer tokens. Code execution with the Model Context Protocol (MCP): anthropic.com/engineering/co…

Claude Code is scary good... not just for coding My Finder kept freezing after it grew to 8GB in size. I had no idea why. Restarting did not help. Turning off Spotlight indexing didn't help. So I turned to Claude Code... and ~10 iterations+30 minutes later it found the issue:

also claude code sdk is criminally underrated and it might be cool to do a cookbook or something for it. i think biggest issue is just lack of resources beyond the docs, which are good, but needs more "how to" material. imo the most underrated piece of software in the world rn.

In our investigation, we uncovered three separate bugs. They were partly overlapping, making diagnosis even trickier. We've now resolved all three bugs and written a technical report on what happened, which you can find here: anthropic.com/engineering/a-…

We're giving Pro and Max users free usage credits for Claude Code on the web. Since launching, your feedback has been invaluable for improving Claude Code. We’re temporarily adding free usage so you can more flexibly experiment with the product.

Use Claude Code as your Video Editor Believe it or not, every video you see below was vibe coded by Claude Code. 🧵 here’s how you can do it too:

GOOD MORNING NEW YORK CITY COME DO YOUR BEST THINKING AT OUR THINKING SPACE IN THE WEST VILLAGE SAY NO TO SLOP

"Our best features come from engineers prototyping an idea, shipping it to dogfood, and just listening to feedback." Here's my new episode with @_catwu (Claude Code product lead) where she shared exactly how her AI-native team operates: ✅ Engineers own features end-to-end ✅ Designers commit production code ✅ Demos over Google docs ✅ Rapid feedback loops every 10 minutes Cat also shared her 3 best Claude Code tips and what's next for the product. 📌 Watch now: piped.video/jmHBMtpR36M

Introducing Claude Code on the web. You can now delegate coding tasks to Claude without opening your terminal.

I have been struggling for months to explain why I think Claude Code is a BIG deal for non-technical people as well. Finally seem to have found a framing that might make sense..

Why even non-coding agents need bash I've done dozens of calls with companies making general agents over the past few weeks and my advice generally boils down to: "use the bash tool more" Here's a concrete example from my email agent:

can yall pls tell me the lore behind your profile pictures

I heard that Shopify and Cloudflare are both increasing their intern intake this year because an intern armed with AI tools can produce value a whole lot faster than interns in previous years Any other companies doing this?

We’re at a fascinating point where there’s a decent thesis for starting new companies from the ground up purely to take full advantage of the leverage you get from AI agents. So much about your typical practices have to be re-engineered to get the greatest output from agents. If you don’t almost completely start your process from scratch you’ll likely cap out early in the gains you can get. We’re seeing this first in the use of AI coding agents, where the workflow becomes far more about clarity of spec writing and prompting than actually writing code. Now the job is far more about the upfront thinking of what you want to build and getting that right, and then reviewing and orchestrating the outputs. This then presents an opportunity not only for new startups that emerge that will build their products in this new way to outrun larger companies, but equally opens up opportunities for new services firms to emerge to bring this engineering leverage to larger customers. The same will be true in almost all other fields. One could imagine new forms of consulting companies focused on getting more output per project, new law firms that can take on far more clients, brand new marketing agencies that can do campaigns entirely faster, and so on. Eventually bigger companies will figure out how to get these gains as well, but for now there’s a clear window for many new firms to emerge that can operate by default in these new ways.

As always Simon is a good way to understand these sort of things, but I think I also increasingly expect shape of "LLM interacts with computers on your behalf" to mean "It writes ~disposable code" versus "It clicks buttons on a web interface, poorly, like Selenium but slower."

Can y’all even name the 3 branches of government?

Introducing Claude Haiku 4.5: our latest small model. Five months ago, Claude Sonnet 4 was state-of-the-art. Today, Haiku 4.5 matches its coding performance at one-third the cost and more than twice the speed.

We launched SWE-Bench Pro last month to incredible feedback, and we’ve now updated the leaderboard with the latest models and no cost caps. SoTA models now break 40% pass rate. Congrats to @Anthropic for sweeping the top spots! 🥇Claude 4.5 Sonnet 🥈Claude 4 Sonnet 🥉Claude 4.5 Haiku