We tried sprints. We removed them.

We ran sprints for six months because that's what the playbook says. We ran without them for six months because we wanted to compare.

The without version is unambiguously better. Throughput is up. Standup quality is up. The thing that went away is the bi-weekly ceremony of pretending the world is predictable two weeks out.

Sprints work in environments where the work is well-specified and the cost of context-switching is high. Most software teams are not in that environment. They're in continuous discovery, and a two-week boundary is a cost, not a tool.

We kept the parts that matter: a regular cadence for review, a clear definition of done, explicit blockers. We dropped the parts that don't: sprint planning, sprint retros, the artificial commitment.