👋 HEY, IT'S RAJA

A break from Apple beta season today because something changed in Claude Cowork that I think is underrated, and the coverage has mostly been feature-list stuff.

Cowork launched in January as a desktop app. Your session lived on one machine and died with it. Close the laptop, and the work stopped. That was always the structural limit, and it's why most people tried Cowork once and went back to chat.

Sessions now run in Anthropic's cloud, which is currently in beta. Close your laptop, and the work continues. Scheduled tasks fire even when no device is online. Your sessions and files are tied to your Claude account and follow you across desktop, web, and phone.

That sounds like a convenience upgrade. It is actually a change in what the tool is for, and there is a catch underneath it that nobody explains properly. Here is the whole thing.

Thousands of AI papers, model releases, and product launches ship every week. You don't need all of them. You need the ten that matter.

TLDR AI is the free daily newsletter curated by Anthropic and ex-Google engineers. People who build AI for a living read everything, then send you what's actually worth your time: the models, research, and tools that will still matter next month.

No hot takes, no hype cycles, no 60-tweet threads. Clear summaries you can read between meetings.

Join 1.1M+ readers. Subscribe for free.

Subscribe for free

What moved to the cloud, what still requires your Mac to stay awake, and how to set up your first scheduled task without it failing quietly.

There is a stat from Anthropic worth sitting with. In an analysis of 1.2 million Cowork sessions, business operations was the largest use case, followed by content creation. Software development was under 9 percent.

Which means the overwhelming majority of people using Cowork are not developers. They are people with recurring work: reports to compile, inboxes to triage, research to gather, documents to produce. That is precisely the kind of work that benefits from running on a schedule rather than being started by hand every time. Until last month, it could not.

This is the part to understand before you set anything up, because getting it wrong is how your first scheduled task quietly does nothing.

The rule of thumb I use: anything a remote assistant could do from their own computer now runs in the cloud. Anything that requires them to be at your desk still needs your Mac to be awake.

Runs in the cloud, no device needed: web research and browsing, writing and drafting, work through your connected services, producing documents, spreadsheets, and slides.

Still needs the desktop app open and your Mac awake: reading and writing files in folders on your computer; anything that uses local connectors; Claude driving your browser as part of a task; and computer use where Claude clicks and types on your screen.

That split is not a shortcoming of Claude. A server in someone else's data center has no route to your laptop's file system. Every cloud automation tool hits the same wall. But it does mean the design question for every task you build is: does this touch my machine, or not?

Scheduled tasks are available on all paid plans, including Pro, Max, Team, and Enterprise.

There is also a Create with Claude option that opens a new task pre-filled with a prompt asking Claude to build the scheduled task for you. Faster and honestly better than most people's first attempt at writing one from scratch.

Each scheduled task runs as its own session, and you review the results the same way you would any other task. Upcoming and past runs are both under Scheduled in the sidebar.

1. The Monday research scan. A weekly task that reads through the Apple and AI coverage from the past seven days and comes back with a short list of what actually changed, separated from what was just rumor. Runs Sunday night, waiting for me Monday morning. This one replaced about ninety minutes of tab-hoarding.

2. The Friday numbers compile. Pulls the week's figures into a single document in the same format every time, so I can compare like with like instead of rebuilding a view each week. The value here is not the time saved; it is that it happens even in the weeks I would have skipped it.

3. The follow-up list. Reads through the threads that have gone quiet and comes back with a list of who is waiting on me. It drafts nothing and sends nothing. It just tells me where I have gone silent on someone, which is the part I am worst at noticing.

The pattern in all three: they compile and report; they do not act. That is where I would start too. Let it prove itself on read-only work before you hand it anything that sends, publishes, or deletes.

Cloud sessions are in beta, and beta means beta. Some features are not yet available on every surface, and the mobile experience is strongest for starting a task, checking status, steering mid-task, and picking up results, rather than for doing real work on a phone screen.

The genuine friction is the local split. If your working life lives in folders on your Mac, a good chunk of what you would most like to automate is exactly the part that still requires the machine to be awake. Worth knowing before you build a week's worth of tasks around a promise that does not apply to your setup.

Start with one task. Make it read-only. Give it a fortnight before you add a second.

See you Friday with five useful things, and this week they are about a panel on your Mac that has been in the security news for a fortnight.

Keep it useful,

Raja · The Useful Tech

P.S. Enjoyed this? Forward it to a friend who loves their Apple gear a little too much.