Skip to main content

Iterating
is not repeating

Each iteration must incorporate new information. If you return to the same point without learning something new, you're not iterating.

Statement​

An iteration without learning is a cost without return. Each cycle of the process must incorporate new information — from the market, the team, the context — that materially modifies the output. If the result of the iteration is indistinguishable from the previous one, something is broken in the process.

Why it matters​

"Iterate" has become an empty mantra in the industry. Many teams repeat cycles without incorporating real learning, confusing activity with progress. In Problem-Driven AI, each iteration has a contract: it must contribute new information that didn't exist in the previous iteration.

The difference between iteration and repetition is the presence of documented learning. If you can't articulate what you learned in this iteration that you didn't know in the previous one, you didn't iterate.

Practical implications​

  • Document the learning of each iteration before starting the next one.
  • Define entry criteria per iteration: What new information do you have? What changed since last time?
  • Kill sterile iterations: if two consecutive iterations produce no new learning, stop and review the process.
Anti-pattern: The Iteration Theater

What it is. Executing sprints or cycles by inertia, without incorporating new learning. The team follows the ritual of iteration (planning, review, retro) but the output is functionally identical to the previous cycle.

How to detect it. Retrospectives produce the same action items cycle after cycle. The team cannot articulate what they learned in the last iteration that they didn't know before. Sprint reviews showcase features but never reference validated learnings or changed assumptions. The iteration number increases but the understanding doesn't.

How to prevent it. Require a documented learning artifact at the end of every iteration. Define entry criteria for each new iteration: what new information justifies this cycle? Kill sterile iterations: if two consecutive cycles produce no new learning, stop and review the process rather than starting a third.

Connections​