Shop, play a game and compile course info with Operator
Operator played a deepfake game scoring 6/10, shopped for jeans, and compiled information about online courses from the author's website. Task: Leon Furze used OpenAI's Operator, a browser-controlling agent, to perform three tasks: playing a Deepfake game, shopping for specific jeans, and compiling course data from his website. He initiated the tasks via text prompts, monitored the virtual browser's progress, and reviewed the outputs. For the course compilation, he delegated the browsing entirely while he took a break, relying on the agent to extract fees and descriptions.; tools: Operator, Claude computer use. Human role: Leon Furze assigned tasks to Operator, monitored its progress, and reviewed the outcomes, including noting the incorrect size selection for the jeans.; limitations: Operator requires a $200/month Pro subscription and is currently in research preview, meaning it is unstable and prone to errors like incorrect clicks. It operates via screenshots and prompts, which can be slower and less reliable than direct API integrations. Human oversight is mandatory to catch mistakes, such as the wrong-size selection, before finalizing actions..
Tools
Published on Agent Work Atlas:
Overview
Operator played a deepfake game scoring 6/10, shopped for jeans, and compiled information about online courses from the author's website.
Workflow
Leon Furze used OpenAI's Operator, a browser-controlling agent, to perform three tasks: playing a Deepfake game, shopping for specific jeans, and compiling course data from his website. He initiated the tasks via text prompts, monitored the virtual browser's progress, and reviewed the outputs. For the course compilation, he delegated the browsing entirely while he took a break, relying on the agent to extract fees and descriptions.
Reported result
Operator scored 6/10 on the Deepfake game in two minutes. For the jeans, it located the correct item but failed by clicking the wrong size before adding it to the cart. It successfully compiled the requested course list from practicalaistrategies.com, returning accurate fees and descriptions after the user returned from a coffee break.
Human role
Leon Furze assigned tasks to Operator, monitored its progress, and reviewed the outcomes, including noting the incorrect size selection for the jeans.
Limitations
Operator requires a $200/month Pro subscription and is currently in research preview, meaning it is unstable and prone to errors like incorrect clicks. It operates via screenshots and prompts, which can be slower and less reliable than direct API integrations. Human oversight is mandatory to catch mistakes, such as the wrong-size selection, before finalizing actions.
Try it
Act as an AI agent using OpenAI Operator to compile course information. Input: URL 'practicalaistrategies.com' and target fields 'fees' and 'descriptions'. Steps: 1. Navigate to the site. 2. Extract all course titles, fees, and descriptions. 3. Format output as a JSON list. 4. Stop and ask for human confirmation before finalizing. Note: This is a suggested workflow; verify accuracy against the source manually. Do not publish or send data without user approval. If the site structure changes, report the error immediately.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.