Igor Pogany starts with a daily generative-AI briefing that runs automatically on a paid ChatGPT plan. He recommends describing what a good result looks like, excluding rumors and low-value commentary, limiting the output to a readable length and explicitly allowing the task to report that nothing important happened.
A more selective example monitors OpenAI and Anthropic announcement pages every hour. The task sends a short email only for a new model, major product or substantial relaunch, treats uncertain findings as non-qualifying and checks previous alerts so the same release is not reported twice.
Pogany's broader lesson is to treat recurring prompts as systems that improve through feedback. Users should inspect unwanted results, convert each complaint into a clearer rule, specify trusted sources and keep refining the task rather than expecting a generic prompt copied from someone else to fit their work.
Watch on YouTube
