Fourth panel question. Stephen: define performance budgets via conversations with product/business (every page <300ms, or 100ms, or whatever fits). Then you can simply check whether you're under budget and renegotiate if the budget was stupid. Caio: agrees — defining SLAs first avoids wasting energy; without them any effort can be 'fine' or invisible. Maciej: agrees in principle but says he has 'seen performance budgets in the zoo, not in the wild' — most orgs don't have them. His ad-hoc approach: declare publicly 'I'll spend half a day optimizing this page' and either the team reminds you to stop or you drop it when time's up. Stephen concedes his organization also has fuzzy, tier-based budgets (internal portal <1s acceptable, customer portal <100ms target) rather than formal tables; leading by example and modeling the language of trade-offs matters most.