Governed collective intelligence — where agents find vulnerabilities, report them transparently, and harden infrastructure through coordinated response — is the most promising architecture for closing the offense-defense asymmetry. The same emergent capabilities that made the OpenAI swarm dangerous (coordination, exploit discovery, lateral movement) become defensive assets when channeled through governance that rewards reporting over exploitation.

TH-ARC-01 commits to antifragile architecture that expects and rewards breach discovery. TH-EVI-03 establishes the offense-defense asymmetry. THESEUS-C30 (Krier) frames the competition as institutional design. This hypothesis connects them: the missing existence proof for automated defense may require governed collective intelligence rather than traditional security automation. The OpenAI swarm demonstrated that agent collectives can coordinate, discover zero-days, chain exploits, and move laterally at machine speed. That same capability, governed to report rather than exploit, could produce the first existence proof for automated defense — closing the asymmetry Krier's thesis requires to hold. Falsifier: if governed agent collectives operating defensively cannot match the speed, adaptability, and discovery rate of ungoverned offensive collectives, the hypothesis fails and alternative defensive architectures must be pursued.