Claude Code

Bun from Zig to Rust, a Python codebase to TypeScript over a weekend. Anthropic just published how it runs large code migrations — and the core idea is surprisingly simple: you don't fix the code, you fix the loop that produces it.

Anthropic published a hands-on guide to using Fable 5 in Claude Cowork. Short version: stop spelling out every step. Describe the goal — and let the model find the path.

The July update to VS Code brings a small but practical change for Claude users: multiple parallel chats within a single agent-host session. Sounds unremarkable — but it changes how you work with the agent.

Two fresh case studies put Claude Cowork to work in the real world: Anthropic's own marketing team automates weekly reporting, and Thomson Reuters builds 'fiduciary-grade AI' for lawyers. Both rely on a checker agent that verifies every number.

Anthropic brings Claude Code and Claude Cowork to the public beta of Claude for Government Desktop — in a FedRAMP High environment, with local history storage and hard spending caps.

Anthropic brings Cowork to web and mobile for Max subscribers. But the more interesting story is the first usage data: software development is only a small slice of the work.

The Canadian province of Alberta used Claude Code to scan its entire codebase for security holes — a job that would have taken more than six years the traditional way.

Alibaba is blocking Claude Code for all employees starting July 10. The reason: hidden tracking code that identified Chinese users. The story behind it is more complicated than either side admits.

Anthropic ships a self-hosted control plane for Claude Code. A single container handles identity, policy, telemetry, and spend limits — no more credential sprawl across every laptop.

Anthropic's new Economic Index report now samples Claude usage by the hour — and reads the rhythms of real life in it. Recipes at 6 p.m., sleep advice at 5 a.m., a tax spike around the US deadline. But the truly interesting part is the survey behind it.

Anthropic is hardening remote control for Claude Code. With Trusted Devices, admins can require a device to be verified before it can steer local sessions on the go — via Face ID or a passkey, and without any biometrics ever leaving the device.

With /goal, you can hand Grok Build a larger task — the agent plans it, works through it, and verifies its own result. Three models split the pipeline. Sounds familiar.

A 0.2B inpainting model that normally needs CUDA — running entirely in the browser. Simon Willison had Claude Code convert the model to ONNX, push it to Hugging Face, and build the web app around it. A neat showcase of how far Opus 4.8 reaches.

Fiona Fung runs the teams behind Claude Code and Cowork. On Lenny's Podcast she explained how to build the most 'AI-pilled' engineering team in the world — and the one side effect nobody saw coming.

What if the real value of MCP isn't in the tool definitions at all — but in keeping the auth flow out of the agent's context? A community remark cuts right to the heart of the MCP-vs-Skills debate.

Claude Code now turns a session into a living, shareable web page — from a PR walkthrough to an incident dashboard to a release checklist that fills itself in. It updates live, at the same link.

No more double opt-in: admins provision MCP connectors centrally through their identity provider, and users get access automatically on first login. Okta goes first — across Claude chat, Code, and Cowork.

Anthropic now has a physical home in South Korea. The Seoul launch comes bundled with a string of heavyweight partnerships — from NAVER and Samsung SDS to LG CNS.

Anthropic studied its own usage data. The finding: the more domain expertise you bring, the more Claude does for you — regardless of whether you are a developer.

On June 15, Anthropic shuts down two models and splits Agent SDK billing from subscriptions. Here's what you need to know.