The 'campaign' is a relic of TV-era media buys. When attention is continuous and creative is cheap, the unit of work is the loop — not the launch.

The campaign model was invented for a world where media was scarce, expensive, and pre-bought. You committed budget six weeks out, shipped a hero asset, ran it for a window, and waited for the agency to mail you a PDF. That world is gone, but the rituals around it — the kickoff, the brief, the launch date, the wrap report — are still everywhere.

A loop has no start or end. It has a hypothesis, a daily cadence, and a memory. Yesterday's winners get more budget today. Yesterday's losers get killed before lunch. The agent doesn't 'launch a campaign' — it adjusts a system. Reporting stops being a quarterly post-mortem and becomes a live readout you glance at the way you glance at a thermostat.

What changes when you operate this way: (1) creative volume goes up 10–50x because variants are cheap and the loop needs fuel; (2) decision latency drops from weeks to hours, because nobody is waiting for a status call to kill a bad ad; (3) budgets get re-cut daily instead of monthly, so dead spend is measured in hours, not weeks; (4) your reporting stack collapses, because the agent is both the operator and the analyst.

The hardest part isn't the tooling. It's killing the rituals. Stop scheduling launch calls. Stop writing wrap decks nobody reads. Replace them with a single live dashboard and a weekly 30-minute review where the only question is 'what hypothesis are we testing next?' Teams that make that switch report the same thing: less meeting time, more output, and a strange new feeling that marketing is finally a system instead of a series of fire drills.

If you remember one thing: a campaign is a project. A loop is a product. Projects end. Products compound.