mission-commandlisted

# Mission-Command Development ## What this is Mission-Command Development adapts military mission-type command (Auftragstaktik) to AI-driven software development. It answers a practical question: how much should a human specify when delegating work to an AI agent? There are two failure modes: - **Dumping everything on the AI** ("build me an app") with no stated intent or constraints. The AI has nothing to anchor decisions to, so it drifts — architecture wanders, priorities are guessed, and nobody can tell if the result is actually what was wanted. - **Micromanaging every step** (specifying each file, each command, each line). The human becomes the bottleneck, the AI's judgment goes unused, and the throughput advantage of delegating disappears. Militaries solved an analogous problem long ago: give a subordinate **intent and constraints**, and delegate the **means**. That is mission-type command. This suite applies the same shape to AI delegation. ## Maximize delegation, minimize constraints Hand off the entire **How** — implementation, sequencing, verification — to the AI. Do not order the steps; order the outcome. Constraints should be **few in quantity and sharp in quality**. These are independent axes, and you need both: - **Few**: every extra constraint is a place the human has to keep re-litigating, and it crowds out the AI's own judgment. - **Sharp**: an ambiguous constraint is worse than none, because the AI can satisfy its letter while drifting through its spi