Apple researchers taught AI to practise using apps without needing a real app to practise on

Key points

• Apple ML Research published a method in 2025 that trains AI agents to use software tools without requiring a live, working version of that software.
• The technique uses a large language model, the technology behind chatbots like ChatGPT and Claude, to act as a pretend environment that the learning agent practises inside.
• Current approaches to building API-calling agents, software that can operate apps and services on your behalf, depend on real databases full of realistic data, which is slow and expensive to create.
• The new method needs only a written description of an API, short for application programming interface, the set of rules that lets one piece of software talk to another.

Imagine teaching someone to use a cash register without ever giving them a real till. You describe how it works, and a colleague plays the role of the machine, responding as the real thing would. That is roughly what Apple ML Research has done for AI, but the colleague is also an AI.

Building an AI agent that can use software tools, booking a flight, querying a database, sending a message on your behalf, currently requires thousands of worked examples. Each example is a recorded back-and-forth between the agent and an actual running application, complete with realistic fake customers, orders, and history loaded in. Creating that data at scale takes enormous effort.

Apple's approach sidesteps the problem entirely. Given nothing but an API specification, a document describing what a piece of software can do and how to ask it, the system conjures a simulated environment on the fly. A second large language model plays the role of the application, inventing plausible responses in real time. The learning agent practises against this stand-in, building up skills without a single real server involved.

This matters because the bottleneck today is not the AI itself. It is the data pipeline. Researchers must build working test environments, populate them with believable dummy data, and run thousands of simulated tasks before an agent is ready to help a real user. That process limits how quickly new capable agents can be created.

The environment-free method could change the economics of that process. Developers would need only the documentation that software makers already publish, turning a months-long data-collection effort into something far faster.

What does this mean for people who use AI assistants?

For ordinary users, better-trained agents mean fewer frustrating mistakes when an AI tries to book, search, or manage something on your behalf. If agents can be trained cheaply across many different apps, the range of tasks they handle reliably should grow.

The work is early stage research, and Apple ML Research has not announced a product built on it. Real-world performance will need independent testing before anyone can say how well the simulated training transfers to genuine applications. But the underlying idea, using one AI to teach another without needing a live environment, is a clean solution to a genuine, well-documented problem in the field.