Keep vs ceremony

There is one test for every Agile practice in this workflow: does it add a move the Author/Operator loop already needs, or only process around the loop? The test sorts practices into three: keep the ones that drive a loop move; adapt the ones whose intent survives but whose team form is replaced by a one-pair form; drop the ones that exist to coordinate people the pair does not have.

None of this says the dropped practices are wrong. A daily standup keeps a six-person team aligned; velocity helps a manager staff the next quarter; planning poker surfaces disagreement between estimators. They answer real problems — problems a one-agent, one-reviewer pair does not have. Dropping them is not cutting corners; it is removing process that protects against a risk the pair has already designed out.

The payoff of the sort is concrete. Select any set of practices below. The readout reports how many of them drive a loop move and which moves are covered, and how many are dropped ceremony that add count but no move. Turn on only the kept and adapted practices and every selection maps to a stage the loop runs; add a dropped one and the move-count does not rise — it is process the loop does not need.

This is the test made visible: a practice that does not raise the covered-move count is not earning its place. Trim it and nothing the loop needs is lost.

On the Portal — the learning platform this course builds — the web chapter F6 shipped nine rungs with the kept set and none of the dropped one. It keeps short iterations (each rung is a PR-sized increment), acceptance criteria (each rung carries a Given/When/Then definition of done), and continuous feedback (every rung's iteration line names the feedback asked). It adapts the retro into the roadmap's per-rung feedback column, and pairing into the Author/Operator split — the Author writes the spec and builds; the Operator reviews and feeds back.

And it drops the rest. There is no standup, no velocity chart, and no planning poker in the F6 roadmap, because the second estimator, the staffing forecast, and the team sync those practices serve do not exist for a one-agent pair. Nine rungs shipped without them, and nothing the loop needs was missing.

# F6 — the Portal on the web — ran the kept set and dropped the team ceremony.

# KEEP: short iterations, acceptance criteria, continuous feedback — each names a loop move.
id = Portal.ID.generate(:course)   # one PR-sized rung, shipped and demoed
assert Portal.ID.decode(id).type == :course   # the rung's acceptance criterion

# ADAPT: the retro → the roadmap's per-rung feedback column; pairing → the Author/Operator split.
# DROP:  no standup, no velocity, no planning poker — the team they coordinate does not exist.

The keep-versus-ceremony sort runs on one test: does a practice add a move the loop needs, or only process around it? Keep what drives a move — short iterations, acceptance criteria, continuous feedback; adapt what survives in a one-pair form — the retro and pairing; drop what coordinates a team the pair does not have — standups, velocity, poker. Distilled, Agile is the handful of principles that name a loop move, and the workflow that runs them is the Author/Operator loop. That closes A3.1 — the next module takes the kept set into Extreme Programming and small batches.

The practices this dive sorts come from the XP and Continuous Delivery tradition; the pragmatic test — keep what earns its place — from the craft. Where the loop they serve is drawn in the course.