Omar Zenhom argues that bad AI answers usually come from bad inputs, not broken models. Treat AI like a new employee who needs onboarding: give it context about your world, goals, and preferences, and it will work better.
So when people say, "AI gave me a bad answer," most of the time it's actually the input that gave the bad answer, and it's just guessing and giving you an average output based on what you told it. The best way to look at it is like you hired a ridiculously capable assistant. It works instantly, 24/7. It can produce high-level work, but it has zero context about your world. You need to give it context. Who you are and what you want. Basically, you got to train it just like you got to train a new employee.
