Skip to content
Atomic Bible
$5 packet preview · Atom No. 002

Evaluator-Optimizer Review Loops

A pattern where one AI system produces work and another applies a narrow rubric before a human accepts the result.

2026-06-095 minute readPaid archive

Archive preview

Historical essays show the thesis and workflow preview. The full examples, prompt pack, and downloadable packet unlock with a one-time purchase or archive pass.

The signal

Teams are getting better results by separating generation from evaluation.

Why it matters

The useful shift is to make AI work auditable instead of treating a single answer as done.

The evaluator-optimizer loop makes quality explicit.

Workflow

Workflow preview

  1. Define the acceptance rubric before generation.
  2. Generate the first artifact.
  3. Run an evaluator pass against only the rubric.
Prompts

Prompt preview

Review this output only against the rubric below. Return defects, severity, and the smallest repair plan.
Watchouts

Where this goes wrong

  • A vague evaluator becomes a second generator.
  • The loop is weaker if the rubric is written after seeing the draft.
Sources

Sources checked

  • Archive preview. Full packet unlocks the rubric set.
Packet

Evaluator rubric pack

  • Rubric templates
  • Repair prompts
  • Before/after examples
Unlock the full packet