Cutting Work Under a Hard Cost Cap

Recall the /usage retro from the case study: the generalization workflow in Phase 5 (auditing every pagination/export path for the missing-tiebreaker pattern, with adversarial verification) was the single largest line item in the session, by a wide margin, ahead of the three-subagent exploration, the wrong-turn timezone detour, and the implementation itself.

Suppose your organization imposes a hard cap: this entire incident, start to finish, has to cost no more than roughly a third of what the session in the case study actually spent.

1. What gets cut or scaled down first, and why that item specifically rather than another?
2. Propose one concrete way to keep most of the value of the cut item while spending meaningfully less on it.
3. Which phase would you refuse to cut no matter how tight the budget, and justify that refusal in terms of what happens downstream if it is cut.