Sessions that talk to each other, design in the terminal, background computer use and a way to test plugins. Here is what changed and why it matters.
Claude Code is Anthropic’s coding agent. It reads a project, makes changes, runs commands and tests, and explains what it did. Anthropic publishes a weekly digest of what’s new, and the stretch from August into September was busy. We read every entry so you don’t have to.
Most of these changes are for developers. A few of them change how quickly custom software can be built and checked, which matters to anyone paying for it.
Sessions that share notes
In early August, Claude Code sessions on macOS and Linux gained the ability to message each other. One session can pass a finding or a decision to another instead of a person re-explaining it. A week later, “fork mode” became the default in interactive sessions, so Claude can hand a side task to a helper that already knows the whole conversation.
In plain terms: one part of a project can be researched while another is being built, without the two drifting apart.
Design before code
Mid-August brought /design, a research preview that puts an artboard workflow into the terminal and desktop app. Claude drafts editable screens for an interface, someone picks one, and Claude builds the one that was chosen. The same week added a built-in Concise output style that leads with the result and skips the preamble.
For a business owner this is a useful habit to ask for: see the screen before anyone builds it. It is far cheaper to change a drawing than a finished app.
More of the work happens without a laptop open
Several updates point the same way. A machine running remote control now shows up as a device card on your phone, so a session can be started from the Code tab. The desktop app can resume a session that began in the terminal, with its full history. When a usage limit is hit on desktop, a new option retries the interrupted turn once the limit resets.
At the start of September, Claude Fable 5.1 arrived in Claude Code with a one-million-token context window, which means it can keep a very large project in view at once. On Pro and Max plans, computer use in the desktop app started working in the background on macOS, so Claude can click through an app while the person keeps working.
Checking the work, not just doing it
The release we’d point to first is quieter. In September, claude plugin eval arrived: it runs a plugin against a set of test cases, scores the results and compares them with a run that doesn’t use the plugin. In other words, a team can prove an add-on actually helps before relying on it.
Other entries in the same spirit: /skill-doctor shows what each installed skill costs and how often it is used, /diff opens a live panel of every edit as it happens, and a --restricted mode starts a session without command-running tools for testing on shared machines.
The direction is clear: more of the build runs in the background, and more effort goes into proving the result is right.
What we take from it
None of this replaces knowing what to build. A coding agent is fast at producing software and getting better at checking it. It still needs someone who has sat with the people doing the work, knows which step wastes their afternoon and can say no to features nobody will use.
That is where we spend our time. Tools like these shorten the distance between “this is the problem” and “here is a working version you can try”.
Sources
From the DYOL blog. Photo by Gabriel Heinzer on Unsplash.
Keep reading
All posts
Claude Code, explained for people who don’t write code
You may never open a terminal, but coding agents change what custom software costs and how fast you see it. What to know, and what to ask your developer.

Claude Opus, Sonnet and Haiku 5.5: which one does what
Anthropic shipped three new models in three weeks. A plain guide to the differences, and how to choose between them for real business tasks.

AI agents on a schedule: what it means for a small team
Agents can now wake up at 7 a.m., check something and report back. What that looks like in a café, a clinic or a club, and where to be careful.
A free 20-minute conversation, then an honest assessment. You’ll know what it costs and how long it takes, in writing, at a fixed price.
Or email lloyd.delacruz@outlook.com. We reply within 24 hours.
