Claude Just Went From AI Tool to Always-On Work

Anthropic clearly got tired of everyone talking about OpenClaw instead of Claude. Over the past month, they've shipped so many features for Claude Code and Claude Cowork that the updates basically blur together. What's wild is that individually, each feature seems like a nice productivity boost. Together, they represent something weirder: a fundamental reimagining of what it means to work with AI.

The feature dump started in late February with remote control—the ability to start a task on your desktop and pick it up from your phone. Your local machine keeps running the session, which means Claude maintains access to your full environment, file system, and tools. You're basically carrying around a window into your desktop that never closes.

Peter Levels compared it favorably to SSH, which tells you the technical crowd was paying attention. But the more interesting response came from people who weren't thinking about the tech at all. Developer Gage Saluda captured it: "You kick off a task in the terminal, then pick it up from your phone on a walk. That's not a productivity feature. That's a relationship shift. You stop thinking of it as a tool you operate and start thinking of it as something you delegate to and check in with."

When Your Phone Becomes Mission Control

A couple weeks later came Dispatch, which took that concept and pushed it further. Unlike remote control—where you're essentially viewing the same session from different devices—Dispatch is an orchestrator. You have one persistent conversation with Claude on your phone, and from there you can spawn and manage multiple task sessions running simultaneously on your desktop.

Anthropic's Felix Reeseberg described it as "one persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work." The constraint is that your desktop has to stay running, but the payoff is that Claude retains context across everything. It's the same conversation whether you're messaging from the train or sitting at your desk.

Developer Paval Huran spent 48 hours testing it and wrote about the experience. His schedule that day: morning coffee (started competitor analysis and Notion page drafts), dog walk (redirected tasks one-handed), passenger seat with wife driving (reviewed and refined outputs), jump arena with kids (tweaked infographic iterations), back at desk (reviewed and shipped everything). Total direction time: 25 minutes. Claude execution running in parallel: 3+ hours of work.

"Dispatch didn't fill my dead time," he wrote. "It changed how I structured my day. I went to the jump arena with my kid because I could direct work async from the sidelines. The model isn't grind during gaps. It's design your day differently because the work runs without you sitting in front of it."

That's the shift people keep coming back to. These aren't features that make your existing workflow faster. They're features that make you question whether your existing workflow makes sense at all.

Channels, Scheduling, and the Always-On Problem

Then came channels—essentially a way to push external events into your running Claude Code session. Telegram messages, Discord pings, CI failures, monitoring alerts, webhook payloads. Anything that can send a POST request can now reach your active session, and Claude can react without you being at the terminal.

This is where things start getting technical and philosophical at the same time. Anthropic's Tariq positioned channels as "more focused on devs who want something hackable." But the implications go beyond dev workflows. You're creating an environment where Claude is persistently aware of events in your digital ecosystem, responding to triggers you've set up.

Add scheduled tasks—both local and cloud-based—and you've got AI that operates on its own timeline, not just yours. Morning briefings, weekly spreadsheet updates, nightly analysis of CI failures. The computer doesn't need to be a place you go anymore. It's a place where work happens whether you're there or not.

Full Computer Control (Yes, Really)

All of that was prelude. Monday night, Anthropic announced that Claude can now control your computer directly—mouse, keyboard, screen, any app. Within 16 hours, 40 million people had viewed the announcement.

The technical implementation is straightforward: when Claude needs to complete a task and doesn't have a dedicated connector, it just... uses your computer the way you would. Opens apps, navigates browsers, fills spreadsheets. Felix Reeseberg noted that Claude reaches for the most precise tool first (existing connectors to Slack, Google Calendar, etc.), but when those don't exist, it falls back to direct control.

Combined with Dispatch, this becomes genuinely strange. You can tell Claude to check your email every morning while you're on the train, make changes in your IDE, run tests, and put up a PR. Or keep your 3D printing project moving according to your initial plan. The work happens on your actual machine, using your actual applications, while you're physically somewhere else.

Developer Sui Berabo's reaction was typical of the breathless early responses: "Claude can now control your entire computer with one prompt forever... Anthropic is giving us the full desktop agent that uses the actual screen, mouse, and keyboard. Not a sandbox, not a simulation. Your real Jarvis."

But Peter Gustv pointed to something more practical: "This is a big deal for a lot of corporates who have custom crappy apps from 20 or 30 years ago." Legacy software that will never get API access or native AI integration can now be automated anyway, because Claude can just... click through it like a human would.

The Security Question Nobody Wants to Think About

Here's what the breathless announcements mostly glossed over: you're giving an AI direct control of your computer. Full stop. Yes, Claude requires approval before acting in Cowork mode. Yes, everything runs locally so your files stay on your machine. Yes, Anthropic has built in guardrails.

But the architecture they're building is fundamentally about persistent access and autonomous action. An always-on AI that monitors events, spawns tasks, controls applications, and operates on schedules you set once and then forget about. That's incredibly powerful. It's also a significant expansion of the attack surface if something goes wrong.

Ethan Mollick noted that Dispatch "feels far less likely to upload my entire drive to a malware site" compared to OpenClaw, which is both praise and an acknowledgment that these are legitimate concerns. The difference between a productivity revolution and a security nightmare often comes down to implementation details that most users will never see or understand.

What We're Actually Talking About

Box CEO Aaron Levie argued that "computer use and the ability to write and run code on the fly are the ultimate primitives for agents to be able to take on more and more tasks and knowledge work. Most work requires hopping between multiple applications and working with broad sets of data in a workflow."

That's the terrain Anthropic is trying to own: not AI as a chatbot you consult, but AI as a persistent layer that operates across your entire digital environment. The features they've shipped aren't just catching up to OpenClaw—they're defining what comes after the current generation of AI tools.

Whether that's exciting or unsettling probably depends on how much you trust the systems you're delegating to, and how much you've thought about what happens when they're not just tools anymore but always-on partners in your workflow. The technology is clearly here. The question is whether we've figured out what we're actually building.

Zara Chen is Tech & Politics Correspondent for Buzzrag