# Automating the Outer Loop Trades Understanding for Ownership ## Claim Removing the human from an agent's outer loop keeps the human accountable for outcomes while reducing their firsthand understanding of how the system reached them. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds specifically when the human previously read every turn and wrote the next prompt. - Applies less where the human was already reviewing only summaries — there's less understanding left to lose. - Doesn't establish that automation is a bad tradeoff, only that it's a real cost that has to be priced in against the human previously supplying stop authority, memory, and review for free. ## Evidence "One big downside of taking yourself out is that you keep the ownership, but would likely lose the understanding." The thread frames this as the honest accounting side of loop engineering: automating the outer loop means rebuilding, as explicit system components, three things a human in the loop previously provided implicitly. ## Links - [[3. Resources/Permanent Notes/The Maker Should Not Grade Its Own Homework]] — because extends; the checker problem and the understanding problem are two separate costs of the same automation move. - [[3. Resources/MOCs/AI Agent Safety MOC]] — because applies-in ## Tests / falsifiers - Survey operators before/after outer-loop automation on self-reported system understanding vs. ownership of outcomes; a flat or improved understanding score would falsify the tradeoff as stated. - Track whether incident post-mortems on automated loops take longer to diagnose than on human-in-the-loop equivalents. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Literature Notes/Prompt Engineering & Loop Engineering, Clearly Explained (Avi Chawla)]] — states the tradeoff directly.