Why it matters
O'Neill names the behaviors that quietly drain a budget — prompt flooding, agent swarms, and background loops nobody owns. It is an engineering leader arguing that a usage dashboard cannot tell activity apart from work that shipped.
Tokenmaxxing read
The lineage is the warning: both predecessors collapsed once teams learned to inflate them, and a token counter is easier to inflate than a story point. His remedy moves control upstream — write the spec before the spend — so intent is the constraint, not a budget alert.
Source takeaway
Bryan O'Neill (InfoWorld opinion, July 14), a CTO with 30+ years leading engineering teams, argues decades of "productivity theater" keep reaching for whatever is easiest to count. His test: did the AI output ship something that matters to a user?
