OpenAI plans to launch a new agent powered by Artificial Intelligence (AI) in January next year and that executes actions as if it were a person, such as typing a code name or booking a trip in a web browser.
It is an intelligent tool called ‘Operator’ internally and will have a similar operation to the one also known by the code name ‘Project Jarvis’ from Google or the Claude 3.5 Sonnet AI model.
Jarvis, for its part, is designed to work in web browsers and, more specifically, Chrome, so that it can help users in everyday tasks and using Gemini 2.0 to respond to user commands or write in text fields.
Sonnet, for his part, pointed out that he had taught Claude different computer skills and software programs for people, with the aim that the model could move the cursor or click buttons, among other actions.
OpenAI has proposed a similar model, which will be able to use a computer to perform actions as if it were a person, such as booking trips or writing code, as people close to the company have advanced to Bloomberg.
This tool will be made available to some users in January, as part of a preview research and through the OpenAI application programming interface (API) for developers.
Three people related to the firm have also indicated that it has been working on different research projects related to agents and that it has finally chosen to a general purpose tool which will run tasks in a web browser.