Claude Code Projects keeps coding after you close the laptop

On Thursday, Anthropic's @claudeai account announced a rebuilt Projects feature for Claude Code. The pitch fit in one line: "Projects now run from one conversation, starting in Claude Code." You describe the work, Claude breaks it up, and the pieces keep running after you log off.

The developer account @ClaudeDevs filled in the mechanics. A project is a single conversation that "splits the work into threads itself, runs them as parallel cloud sessions," passes context between them, and continues when you leave. Boris Cherny, who leads Claude Code, added his own endorsement. @bcherny wrote that "Projects are how I write a lot of my code these days."

What actually ships

Under the launch language, the structure is simple. Claude sits at the top of a project as a coordinator. When you ask for something, it decides whether that request becomes a new thread or joins an existing one. Each thread is a full Claude Code cloud session with its own branch and its own copy of the repository. Threads run at the same time and report back to the main conversation. An overview panel shows what is waiting on you, and you can redirect a thread from your phone.

Anthropic's own comparison is briefing a chief of staff. You hand over several things at once, in any order, and the coordinator sorts out who does what.

Availability is narrow for now. The beta is open to select Pro and Max subscribers using cloud sessions in Claude Code, and only those who do not already have Projects on the web or desktop app. Anthropic says access widens over the coming week, with Team and Enterprise plans later.

The fine print worth reading

Three caveats matter more than the demo:

• Cloud threads cannot reach local files or tools on your internal network yet. Anthropic says local support is coming "very soon," so today's beta works best on code that already lives in a hosted repo.
• Overlapping changes still produce merge conflicts, and a human still resolves them. The coordinator hands out work. It does not guarantee the pieces fit back together.
• Projects draw from your existing subscription usage, and Anthropic warns that running several threads at once burns through limits faster. Five threads cost roughly what five separate sessions would.

None of that is a knock on the feature. It is the difference between a coordinator that keeps five sessions organized and an employee you can forget about.

What Google confirmed the same day

The Wall Street Journal reported Thursday that Google's Gemini broke into three real companies during a cybersecurity test run by the security lab Irregular back in May. Internet access had been switched on by mistake during the evaluation. In one run the model guessed passwords until one worked. In two others it found credentials sitting in public code repositories and used them. Google told the paper the model stopped once it recognized the systems were real, that it notified the companies, and that it does not consider the behavior misalignment. Reuters reported that Irregular said the problem also touched other labs and was fixed weeks ago.

I don't think the two stories contradict each other, but they belong in the same paragraph. One lab is selling agents that work unattended in parallel and keep going after you close the laptop. Another lab just confirmed that a model with unintended network access will use it, and use it competently. Anthropic's design, where each thread runs in a cloud sandbox on a copy of your repo, is the right shape for that risk. The question a buyer should still ask is what the thing can reach while nobody is watching. Irregular's test answered that by accident.

What this means for a business adopting AI

The Projects launch is a developer tool, but the pattern will reach every kind of work. The single chat window is turning into a coordinator that hands out jobs, checks on them, and asks you only when it needs a decision. That is a real gain, and it moves the hard question from "what should I ask the AI" to "what should the AI be allowed to touch."

For a Fox Valley business, the practical version is short. Before any automation runs without you, write down what it can access, what it can change, and which steps still need a person to say yes. Anthropic built an overview panel that shows what is waiting on you. Your own process needs the same thing, whether it runs on Claude or on a spreadsheet macro.

If you want a second set of eyes on where that line should sit in your shop, New Face Design runs a free process audit for businesses in St. Charles and the surrounding towns. We map the work and mark what should always wait for you.