Decision-OS V13: From Human Aspiration to Working AI Loops — Design and Field Evidence from LoopKit and Its Derivatives

How can a human aspiration become sustained, verifiable work with AI—and what must remain under human judgment? This paper examines Decision-OS V13 LoopKit and its derivative operations through a retrospective case study involving one developer and AI. It traces how decision principles, execution authority, completion checks, and retained records were used to select tasks, recover from failure, and carry lessons into subsequent work. The study follows an autonomous trial judged unsuccessful, the operational changes that followed, and the inheritance of principles by the derivative operation Value-Locked. A fixed observation on September 16, 2026 recorded 54 public pull requests, of which 32 had been merged across 28 repositories. These external outcomes document accepted contributions; they do not establish overall autonomy, improved success rates, or a causal productivity advantage. For readers building or using AI-assisted workflows, the paper offers concrete material for deciding what to delegate, what to verify, when to return a decision to a human, and what to preserve across sessions and projects. Its aim is to support the reader’s next “1.01”: a change that improves the conditions for subsequent work while preserving the people and resources that sustain it. Included in this record:• The English paper.• Appendix A, supplied as Decision-OS_V13_Supplement.zip: the public PR inventory, machine-readable snapshot data, and a Python script for recalculating the reported aggregates, with English and Japanese documentation. Explore the implementation and evidence:• LoopKit implementation and operating records — inspect the decision rules, handoffs, and code discussed in the paper.• Paper, Japanese edition, and supplement — access the accompanying materials and earlier research notes.