anti-hivemind-prompt-rewriter

To rewrite prompts and avoid generic LLM outputs, you need to apply explicit diversity constraints and a diverge-converge workflow. This process generates multiple distinct prompt variants organized into clusters based on different underlying assumptions, effectively preventing mode-collapse while preserving your original intent.

The best way to generate diverse prompt variants for product tasks is to parse your initial prompt for constraints and produce 4 or more distinct variants. Each variant should be grounded with explicit assumptions, tradeoffs, and evaluation rubrics to facilitate objective comparison within a diverge-converge workflow.

Yes, you can use a diverge-converge workflow to compare rewritten prompt variants. The process generates multiple constraint-aware variants organized into clusters, then grounds them with explicit assumptions and evaluation rubrics to help you objectively compare and pick the best variant for your context.

Your LLM output looks the same because of dominant-mode behavior, where minor wording changes still trigger mode-collapse. Rewriting prompts with explicit diversity constraints and multiple distinct underlying assumptions forces the LLM to diverge from generic patterns and produce varied, constraint-aware outputs.

You should generate at least 4 distinct prompt variants to ensure output diversity. These variants must be organized into clusters, with each cluster based on a different underlying assumption and mechanism, and grounded with evaluation rubrics to rigorously test output variety.

Yes, prompt rewriting for diversity works for coding and personal-growth tasks. The approach applies explicit diversity constraints and a diverge-converge workflow to any domain where variety and rigor matter, preserving your original intent while generating multiple distinct, constraint-aware variants.