16-minds summons agents modeled on 16 distinct personality types — each with its own values, thinking patterns, and tone — to debate, review, or advise on any topic. Each personality is written directly into an independent prompt, so outputs carry visibly different voices instead of the averaged answer a single agent tends to produce. Three commands cover the range: - /mind <type> — a single type's opinion - /pair <a> <b> — a two-type debate (open → rebut → reply / synthesize) - /minds — fan out to all 16 in parallel, then synthesize Three modes (light / middle / heavy) trade depth for token cost. The /minds synthesis follows a fixed 4-section rubric — score distribution, axes of disagreement, easily-overlooked perspectives, and landing-zone candidates with per-candidate "types likely to agree / fit / risk" — to keep 16-voice output structured rather than scattershot. The --save flag writes the full execution log to a file for use as article material or decision logs. Built for decisions where a single averaged answer would lose the edges: idea generation, design judgment, code review, and life/career questions.