I consolidated Decision-OS V4–V8 into a single gateway (Zenodo DOIs + SSOT). Key idea in V8: evaluate/control *trajectories* (Time-Tube), not point outputs. Self-Safe = harm / collapse / dependency gradient (+ reversibility interventions) AGI論文を書きました。良かったら読んでください。 Gateway: https://github.com/shin4141/decision-os-paper/blob/main/README.md
- 4 replies
- 4 recasts
- 13 reactions
The AI said “done.” Why are you still doing the cleanup? I built Output Surface Integrity (OSI) to make AI completion inspectable: • the exact output • evidence and unresolved items • rollback or restart point • the next safe action and owner Start with one AI output. No formal benchmark required. The repo also preserves adverse results and a blocked Formal Run instead of rewriting them into success. https://github.com/shin4141/output-surface-integrity
- 0 replies
- 2 recasts
- 3 reactions
AI agents do not only need better acceleration. They need brakes, handoffs, and loop gates. I built Decision-OS V13 LoopKit to answer one question before the next AI loop runs: Should this continue, hold, cap, or block? Because faster loops without governance just compound hidden debt. https://github.com/shin4141/decision-os-v13-loopkit
- 0 replies
- 1 recast
- 4 reactions