We are entering a very delicate stage:

AI has not been fully understood, yet it has already been widely authorized.

It is allowed to help you trade, allocate funds, execute strategies,
but in most systems, the cost of errors is almost zero—mistakes are just "regenerated."

From an engineering perspective, this is actually very dangerous.
Because when a system doesn't have to be responsible for errors, what you get is always just "seems reasonable."

This is also why I am more optimistic about paths like @miranetwork.
It doesn't focus on stacking smarter models, but rather on embedding "verification" and "responsibility" into the system's core:
Mistakes are detected, and deception cannot go unnoticed, so costs must be paid.

When AI starts making decisions on behalf of people,
Trustworthiness is not about emotions, but mechanisms.