Ask AI: "How many R's are in the word strawberry?" For a long time, most AI models got this wrong, confidently answering "two" when the answer is three. But why? You can see the letters right there. What makes this simple question so hard for a system that can write entire novels?

The answer reveals something fascinating about how AI "sees" language. It doesn't read words letter by letter the way you do. It breaks text into chunks called tokens, and sometimes a single letter gets merged into a bigger chunk, becoming invisible.

It's like trying to count the number of eggs in a cake after it's been baked. The ingredients have been transformed into something new. Understanding this changes everything about how you talk to AI.