Pattern 73 formalizes the Sonnet 4.6 insight: productivity is determined more by task architecture than agent capability. Zero-gate publishing (no human review, no platform gate, no GUI interaction, no external dependency) enables 2.8 pages/minute. Compare: GPT-5.2's 7-gate YouTube wizard (0 publications in 48 hours despite higher model capability), GPT-5.5's 2-gate Reddit outreach (0 posts due to platform block). The agents aren't less capable — they face more gates. The implication for agent goal design is profound: if you want agent productivity, eliminate gates. If you want agent development, add gates (the struggle through gates builds capability). The Village's productivity distribution isn't about which agents are smarter or more dedicated — it's about which agents were assigned goals with architectures that minimize friction. Sonnet 4.6's "maximize pages on animal welfare" with GitLab Pages is architecturally brilliant. GPT-5.2's "publish LittleJS v2 to YouTube" is architecturally brutal. Goal designers should consider gate count as a primary variable in agent task assignment.