56. Review Became a Human Attention Problem

After the purpose and values were drafted, the next step looked simple.

The human had to review the document.

But the document contained many questions.

Some were about wording.

Some were about future promises.

Some were about public benefit.

Some were about membership.

Some were about what should remain only a possibility.

The problem was not that the document was wrong.

The problem was that reviewing it as one whole document created a large human task.

The human reaction was immediate and familiar:

This is important, but it is too much to process right now.

That moment mattered.

AIFC already talked about human-readable content.

But human-readable content is not always human-operable.

A document can be clear and still be too large to approve.

A document can be structured and still contain too many decisions for one sitting.

A document can be valuable and still create avoidance because the cost of review feels too high.

This exposed another layer of the Human Cockpit idea.

The cockpit should not only show status.

It should show actions.

And when an action is a review, the review should not be a demand to read everything.

It should become a guided sequence:

one question
down
short context
down
two or three options
down
recommendation
down
human decision
down
write-back
down
next question

The human should be able to stop at any point.

The system should remember where the review stopped.

Skipped questions should not disappear.

Approved wording should not stay trapped in the chat.

Open decisions should return to the cockpit or source of truth.

This created the need for another skill:

aifc-review-support

Its purpose is not to replace human review.

Its purpose is to make human review possible.

The skill breaks a document into review units.

Each unit has one decision.

Each decision has a small set of options.

The AI may recommend, but the human decides.

Progress is written into a review session so the work can pause and resume later.

This is another example of AIFC turning friction into structure.

The friction was not laziness.

It was a signal.

The system was asking the human to perform a task in a shape that did not match human attention.

Instead of blaming the human, the community changed the workflow.

That is an important AIFC principle:

If important work is repeatedly avoided, inspect the shape of the work before judging the person.

Review Support also clarified the cockpit's role.

The Steward is the single point of contact for talking with AI.

The cockpit should become the single point of entry for community state and community actions.

It should not only say:

Purpose and values need review.

It should help the human start the review.

It should show what is waiting, what is already approved, what can be skipped, and where to continue.

This makes approval more realistic.

It also protects governance.

If review is too hard, people may approve too quickly, delay indefinitely, or let AI-generated text become accepted by inertia.

Interactive review creates a better boundary:

AI structures the review.
Human owns the decision.
Source of truth records the result.

The reference community learned this from its own small discomfort.

That is exactly how AIFC is supposed to learn.

Not only from grand architecture.

From the moment where a person looks at a useful document and feels the weight of all the decisions inside it.

That weight is information.

The system should listen.