Individuals are using LLMs way more effectively than organizations today.

Individuals are using LLMs in situated contexts, informally and as augmentation to their work, not automation for it.

When an individual uses LLMs, there's a human in the loop every time.

The human in the loop can handle failure cases, and steer it more effectively.

When an organization wants to use LLMs it has to have some kind of automated process or structure; hard, inflexible, less forgiving.

When individuals use AI in their day to day (e.g. employees or students) it's likely something they don't want their boss to know about.

This means that most successful AI use today is subterranean.

It is illegible to the surrounding organization… and the employee has an incentive to keep it illegible.