If you haven’t read the introduction post, that’s a good place to start.
This is the third quadrant in the Learning Terrain Model: fast feedback in a complex domain. It carries a tension the earlier quadrants didn’t. The feedback is still fast — you can iterate often and cheaply — but the domain is now complex rather than complicated. The rules aren’t stable. The same input can produce different outputs depending on state, timing, or what else is running. Behavior emerges from interactions you can’t fully enumerate. You can’t drill your way to mastery, because the thing you’re drilling against keeps moving.
I’ll use the exploratory software tester as the example, partly because it’s the terrain I know best. Picture her sitting down with a new build. A feature that worked yesterday behaves differently today, not because she was wrong yesterday but because the ground moved under her. A bug shows up only when two actions happen in a particular order. Another vanishes the moment she adds a second user. None of this is unmapped-but-fixed territory she’ll eventually pin down for good. A new build is, in a real sense, a new experience — yesterday’s truth has a short shelf life here.
For working this kind of terrain, James Bach’s Session Based Test Management offers a structure that fits almost exactly. It wasn’t built as a learning framework, but its logic maps cleanly: how do you extract real signal from cheap, plentiful iterations when you can’t trust that what held yesterday still holds today?
It starts with a charter — a short statement of what she’s exploring and why. Early on it can be deliberately broad: how does this part of the system respond when I push on it? As her understanding sharpens, the charter narrows. It gives direction without deciding where she ends up.
Then comes the session: a time-boxed burst of direct engagement with the system. The goal is observation and good questions, not conclusions. This is worth sitting with. In a domain that can shift on you, the urge to explain what you’re seeing is exactly the move to resist. Write down what happened first. Interpretation comes later.
Afterward, she debriefs. Before committing to an explanation of what she saw, she names the lens she’s looking through, then forces out two or three alternative explanations for the same behavior. The point is to falsify her own emerging conclusions before they harden. A belief formed too early in a complex domain is a liability — it quietly filters out the contradictory signal that arrives next week. The debrief shapes the next charter, and the loop continues.
Strip the testing vocabulary and what’s left transfers to anything in this quadrant. The charter is committing to a question instead of a plan, because a plan assumes a stability the terrain doesn’t have. The session is giving yourself a window to collect raw signal before you start explaining it. The debrief is refusing to trust your first explanation until you’ve tried to break it. And the loop sharpens the question rather than the answer. A comedian working an unfamiliar room runs the same cycle — try the bit, watch what the room actually does, resist the easy story that they just didn’t get it, adjust, go again. So does a salesperson feeling out a prospect who keeps moving the goalposts. The domain changes; the discipline doesn’t.
What accumulates over time isn’t a rulebook but a pattern library, held with calibrated uncertainty about which patterns hold and which might shift. Some do hold: when you find and verify genuine determinism in a corner of the domain, that corner stops being complex and becomes complicated, and the terrain shrinks accordingly. The end state here isn’t mastery in the pt.2 sense of effortless automaticity. It’s closer to a working map of the territory — knowing which parts you can lean on, which parts still need probing, and being honest about the difference.
