Entire methodology
M5
Thinking & Facture
Over-explanation of basics
Detailed decoding of simple terms and abbreviations, even if the text is clearly intended for narrow professionals.
Neural network (Hyper-explanation)
We updated the CSS (Cascading Style Sheets, responsible for the visual look of the site), after which we had to clear the cache (temporary browser data).
Human (Blind spots)
We pushed new styles, the layout broke. Had to flush the cache.
Why does this happen?
The basic firmware of any commercial LLM is Helpfulness. During the training phase, the algorithm is literally penalized if its answer might seem unclear or incomplete to the user.
When a live person writes a text for their audience (for example, in a work chat or a professional blog), they use the asymmetry of knowledge. They know that they don't need to explain basic things to the reader. A neural network has no understanding of "its own crowd". It always writes like a dictionary announcer for sixth graders.