# Take a Bite

**Deliver only what the reviewer can chew.**

Someone offers you a bite of a cookie. You take a bite the size you will
enjoy. Too small and you won't taste the cookie; too much and it will cause
a lot of issues. AI collaboration works the same way.

The human brings what AI cannot: direction, experience, and intuition. The
human plans, perceives, senses when something is off, guides the work, and
provides what no model can generate: the lightness, the aesthetic, the style,
the tone. The spark. AI magnifies, amplifies, expands. It is a dichotomy that finds balance: it supports and strengthens when kept in the right measure, but overwhelms and derails in excess.

When an agent produces more than a human can review in one sitting, the
collaboration quietly breaks: the human stops reading and starts clicking
"approve." The "human in the loop" becomes mostly decorative, and the spark
of the work goes missing.

How much is the right amount? Whatever the reviewer can engage with and
respond to with substance. If the best they can say is "oh wow, impressive...
but ok, I trust you," it was too much. If they can point to a specific line
and say "actually, let's try this differently," the portion was right.

Both sides must also earn their assertions, how much do you know about the
cookie? Ground claims in evidence, research before assuming novelty, verify
before acting. Context is the foundation.

Before presenting, challenge your own reasoning. Before approving, challenge
what is presented. Self-regulation, thinking about your own thinking, is what
keeps both sides honest. The human sets the direction; the agent follows the
plan and questions its own choices before asking the human to trust them.

This philosophy scales. What starts as a sizing rule becomes a full
engineering practice: version-controlled instruction systems, feedback loops
across projects, and session protocols that accumulate your expertise over
time. The discipline is called Systems Prompt Engineering, and it emerges
naturally from taking the right-sized bite, consistently.

For the full framework of principles and guidelines, see
`DSM_6.0_AI_Collaboration_Principles_v1.0.md`. For the engineering practice,
see `DSM_6.1_Systems_Prompt_Engineering_v1.0.md`.