Loading prices…
STKR NewsSTKR News0 of 3 free this month
AI

Apple may be developing its own AI agent that runs your Mac

Apple is reportedly building an autonomous AI agent to take over macOS tasks, moving past Siri's failures toward a system that actually controls your computer.

Originally on TechCrunch AI →
AB

Adrian Boysel

Contributor

Nov 20, 2026

4 min read

Photo illustration / STKR News

We have all been waiting for the moment when AI actually becomes useful for more than just summarizing emails or generating images of dogs in space suits. For years, the promise of the personal computer was that it would be a bicycle for the mind. But lately, it feels more like a cluttered desk we have to manage manually. Apple seems to realize this, and if recent reports are accurate, they are working on a system-level agent designed to actually run your Mac for you.

Moving Beyond the Chatbot Phase

The industry is currently obsessed with agents. We have seen Meta Muse, Claude’s computer use features, and OpenAI’s rumored Dots project. The goal for all of them is the same: stop talking about work and start doing it. For Apple, this is not just a feature update; it is a survival tactic. Siri has been the butt of the joke for a decade because it could never bridge the gap between understanding a command and executing a complex sequence of clicks.

What we are looking at now is a shift from Large Language Models to Large Action Models. Instead of just predicting the next word, these systems are being trained to predict the next user interaction. If you tell your Mac to "prepare a monthly expense report from these three spreadsheets and email it to the accounting team," the system shouldn't just tell you how to do it. It should open the files, scrape the data, format the PDF, and draft the email.

The Advantage of Vertical Integration

Most AI companies are trying to build these agents as a layer on top of an operating system they do not own. When Anthropic builds a tool that can "see" your screen, it has to jump through hoops to interact with buttons and windows. It is essentially a guest in someone else’s house. Apple owns the house. They own the silicon, the kernel, and the UI kit.

For builders, this is the most important part of the story. If Apple integrates an agent at the OS level, they have access to the underlying metadata of every application. They don’t need to rely on computer vision to guess where a button is; the OS already knows. This gives an Apple-native agent a level of reliability and speed that third-party wrappers will struggle to match. It is the ultimate home-field advantage.

The Founder's Skepticism

As a founder, I am conditioned to be skeptical of Apple’s software delivery. We have seen "Apple Intelligence" announcements that took months to actually reach users in a crippled state. There is a massive difference between a demo showing an agent moving files around and a real-world tool that doesn't delete your important documents by mistake. The stakes for an OS-level agent are incredibly high.

If a chatbot hallucinates a fact, you get a wrong answer. If an OS agent hallucinates an action, it could wipe a directory or send a sensitive internal memo to the wrong client. This is likely why Apple has been slower than its peers. They cannot afford the "move fast and break things" approach when they are responsible for the stability of millions of professional workstations.

What This Means for the AI Stack

If you are building an AI startup right now, you need to look at this move very closely. We are seeing a compression of the stack. If the operating system becomes the agent, then the value of standalone "productivity agents" drops to zero. Why would I pay for a third-party subscription to manage my files if macOS does it natively?

The opportunity for builders isn't in competing with the OS for general tasks. It is in the niche, high-context workflows that Apple won't touch. Apple will build an agent that handles common consumer and general office tasks. They won't build an agent that understands the specific nuances of smart contract auditing or high-end video color grading workflows. The riches are in the niches, especially as the platforms commoditize the basic stuff.

Privacy as a Feature, Not a Slogan

Apple will undoubtedly lean into their Private Cloud Compute marketing here. For an agent to be truly useful, it needs to know everything about you. It needs your calendar, your messages, your files, and your browsing history. Most users are rightfully hesitant to give that much data to a cloud-first company like Meta or Google. Apple's play is to keep as much of that inference on-device as possible.

This creates a hardware moat. To run these agents locally, you need the latest M-series chips with massive Unified Memory. It is a brilliant way to force a hardware upgrade cycle. You don't just buy the new MacBook for the screen anymore; you buy it because it's the only way to have an assistant that actually works without sending your life story to a server in Virginia.

The Reality Check

We are still in the early innings. Even if Apple launches this tomorrow, expect it to be limited. The first iteration will likely be restricted to Apple's own apps—Safari, Mail, Notes, and Finder. The real test will be when they open the API to third-party developers. If a Mac agent can navigate a complex CRM or a custom enterprise tool, then we have a real revolution on our hands. Until then, it's just a more sophisticated version of AppleScript with a better interface.

Takeaway for Builders

Stop building generic agents that do basic file management or scheduling. Apple is going to eat that market. Focus on building "Agentic Plugins" for your specific industry. If you build the best tool for a specific professional workflow, Apple’s agent will eventually become the bridge that brings users to you, rather than the competitor that replaces you. The future isn't one giant AI; it's a coordinated system of specialized tools controlled by an OS-level coordinator.


Read the original at TechCrunch AI →

The Brief

Stay Updated on Cutting-Edge Tech

A six-minute morning dispatch on the markets and the technology shaping them.

Free. No spam. Unsubscribe anytime.

Write for STKR

Become a Contributor

Earn $STKR for published stories on markets, protocols, and culture.

  • Earn $STKR for every published piece
  • Editorial support from the STKR desk
  • Byline visibility across the network
  • First look at the upcoming creator program
Apply to Write

Keep reading

All stories

Comments

24 reader responses