← The Vault
The Big Story

A new 'agentic' AI tool arrives for your smartphone

A new piece of software called OpenClaw has launched for mobile phones. Unlike standard AI chatbots that only write text, this program is designed as an 'agent'—a system that can actually perform tasks for you like clicking buttons or navigating menus, rather than just talking about it. Here is how these programs function and why putting them on your phone changes how you interact with your device.

Edition № 144Room: The Big Story30 June 20262 min readSources: 1
Article

You are used to AI that talks back, but a new wave of software is moving out of chat boxes and into the actual controls of your phone.

WHAT'S HAPPENING

A project called OpenClaw has released its software for iPhones and Android devices. It is an 'agentic' program, which means it is designed to take action on your behalf. Standard AI assistants are like digital librarians that answer questions about facts. An 'agent' acts more like a personal intern; you give it a goal, like finding a flight or checking a status, and it interacts with your phone's apps to get that task done by clicking buttons and navigating menus for you.

The shift from talking to doing

HOW IT WORKS

Most AI currently running on your phone relies on a 'model'—a complex file containing a massive map of language patterns created during a learning process called 'training.' During training, developers feed the system millions of examples of how humans write and think, allowing it to predict the most likely next word in a sentence. Agentic software uses these models to 'reason' about what to do next. It continuously takes a 'snapshot' of what is on your screen, compares that image to your goal, and uses its training to decide which icon or button to tap. It is a loop of observing the screen, deciding the next step, and triggering a command until the task is complete.

WHY IT MATTERS

Putting this capability on a smartphone shifts the AI’s role from a source of information to a driver of activity. Up until now, AI has been an extra layer of reading or writing work for you to review. With agentic software, the program takes control of the interface. This creates new considerations for security and privacy, as these tools require permission to navigate your apps and access your data. As these systems become more common, the challenge will be ensuring they only perform the specific tasks you intended, rather than taking actions the software thinks are helpful but you never actually requested.

Sources
← PreviousAnthropic's New Research Tool and Evolving AI Export RulesNext →The Hidden Energy Drain of Our New AI Habit
Tomorrow's edition · free

Liked this one? The next lands at breakfast.

Every story in tomorrow's AI news, rebuilt in plain English — five minutes, sources linked, free forever.

By joining you agree to receive Article's daily newsletter — unsubscribe in one click. Privacy

← Back to the Vault