Asked four different AIs to name a random number between 1 and 25.
ChatGPT β 17
Claude β 17
Grok β 17
Gemini β 17
LLMs don't generate randomness, they predict what a person would answer.
And it turns out that for some reason people often name exactly 17.
If you use AI without customization and your own data, you get exactly what everyone else gets
Same model -> same answers
The real benefit is when you contribute something of your own: drafts, tasks, correspondence, thoughts, etc.
When a model knows how you think, what is important to you, how you write and what decisions you make, she stops being just an assistant and becomes an extension of your head.
Otherwise, it's just a general template in a pretty wrapper.
