Cua argues that useful computer agents should move between code and visual applications according to the task. Its January 2026 retrospective points to file handling and data processing as work suited to commands, while visual editing and unfamiliar interfaces still give agents reasons to inspect and operate a screen.
The consequence is a broader design for automation. A workflow spanning applications can use code where direct computation is appropriate and desktop interaction where the interface supplies the needed controls or visual feedback. This is Cua's argument for combining capabilities within one AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →, rather than treating every task as a sequence of mouse clicks.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
