OpenAI product manager Ty Geri says one of his go-to instructions for ChatGPT is simple: “Please impress me.” His point, however, is that users should not need a perfectly crafted prompt to get something useful. In his view, the model should already be trying to deliver responses that feel smart, helpful, and a step beyond the obvious.
According to Business Insider, Geri also uses long, wandering inputs with ChatGPT Work to figure out what is “sticking.” That suggests a more natural style of prompting, where users do not have to trim every thought into a polished command before they start. Instead, they can give the system a fuller stream of ideas and see which parts produce the most valuable output.
The approach highlights a broader shift in how some AI insiders think about prompts. Rather than treating prompting as a rigid technical skill, Geri’s advice points toward experimentation and everyday language. If the model is built well, it should be able to handle messy context, pick up the important signals, and return something stronger than a basic summary.
For people using ChatGPT at work, Geri’s comments reinforce the idea that prompting can be less about finding a magic formula and more about exploring what the tool can do with real, imperfect input. His “impress me” habit is a concise way of raising the bar while still expecting the model to do more of the heavy lifting.
