Variance first, constraints last. If you do not start with garbage, you never find the gold.
Once you accept that the Mega-Prompt is a dead end, the real question becomes structural: what replaces one-step generation?
Most production teams optimize for being right the first time. That works for deterministic systems, not probabilistic ones.
Large language models sample. Asking for correctness, compliance, and creativity in one pass collapses output toward the safest answer.
- Fully formatted.
- Fully compliant.
- Completely lifeless.
Step 1: Maximize stochasticity
Creativity has to precede correctness. Early constraints usually suppress exploration.
Increase temperature. Remove heavy rules. Encourage speculation.
The goal is not quality yet. The goal is variance.
Step 2: Human as director
Generate distinct directions, not tiny edits to one draft. The human chooses the best direction instead of repairing mediocre prose.
- Straightforward and clear.
- Opinionated and sharp.
- Unexpected and playful.
Selection replaces correction. Judgment replaces repair. That one shift removes a surprising amount of friction.
Step 3: Deterministic filtering
Only after direction is chosen should hard constraints return: formatting, policy, structure, platform requirements.
By moving rules to the end, they become a safety net instead of a cage.
Core design principle: variance must be created before it can be constrained.
Re-sequencing fixes lifeless output, but tone consistency still needs a stronger method. That is what article three covers.