What ordered unsticking drill do you run aloud when an approach stalls mid-interview?

## The behaviour being tested Everyone gets stuck. Interviewers pick problems that make it likely, because how you *recover* discriminates far better than whether you needed to. The two bad outcomes are recognisable from across the room: silent thrashing (typing, deleting, typing, saying nothing for four minutes) and paralysis (staring, apologising, waiting for rescue). Both produce the same debrief line — *did not have a way forward* — regardless of how strong the candidate is on a good day. The antidote is a rehearsed drill, run out loud, in a fixed order. Fixed order matters because it removes decision-making at precisely the moment your decision-making is compromised. And saying which rung you are on turns being stuck into a visible, competent process rather than an absence. ## Rung one: re-work a small concrete example, by hand Step back to a case small enough to hold in your head — three or four entries on a shared calendar, chosen so the interesting situation actually occurs — and process it manually, narrating each step. Not "what would my algorithm do", but "what do *I* do". This works because human intuition solves small instances using steps you never wrote down. You will catch yourself doing something — glancing ahead, mentally arranging the entries, remembering something from earlier — and that unrecorded step is usually the missing piece of the plan. Say what you notice: "I just compared these two without checking the first one at all — why did I feel allowed to skip it?" That sentence has broken more stalls than any amount of staring at a half-written function. It is also cheap and completely safe to narrate: even if it yields nothing, the interviewer sees concrete, grounded work rather than a freeze. ## Rung two: simplify the problem, deliberately and out loud Solve a version you *can* solve, and be explicit that you are doing it: "I'm going to solve this for a single shared room first, get that right, and then ask what changes when there are many." Good simplifications remove one dimension of difficulty: - collapse a general case to a fixed one (many rooms to one room), - shrink the input to a size where brute force is obviously fine, - assume the input arrives in a convenient form and see whether that assumption alone unlocks it, - drop a requirement (ignore the reporting part, just answer yes/no) and add it back later. The last variant is quietly powerful, because the *cost* of the assumption you made is often exactly the sub-problem you were stuck on — and now you know its name. The explicit second step is what makes this a drill rather than a dodge: after solving the easy version, ask out loud which parts generalise and which break. Interviewers are generally happy to let you take this route; what they dislike is a candidate who solves a simplified version and quietly presents it as the answer. ## Rung three: scan technique families aloud Only now go to the catalogue, and make the scan audible with a one-line verdict per entry rather than a silent flip-through. A workable short list: put the input in a useful order; trade memory for fast lookup; move two coordinated positions through the data; recurse and remember sub-results; keep a running extreme as you go; model it as nodes and edges and traverse. The verdicts are the value: "ordering — plausible, the difficulty is comparing things that arrive unsorted, so that could help." "Two coordinated positions — needs a linear structure with a consistent direction, which I do not obviously have." "Graph traversal — I don't see relationships between entries beyond overlap, so probably not." Even the rejections are signal, because they show you know what each family *needs* in order to apply. A candidate who says why four techniques do not fit is demonstrating more than one who guesses right once. The order matters. Pattern-matching first is the common mistake: it is the most comfortable rung and the least grounded, and it tends to produce a technique bolted onto a problem you have not yet understood. Examples first, then simplification, then catalogue. ## Timeboxing and the exit The ladder is not infinite. Give it a few minutes; if none of the rungs move you, take the exit deliberately rather than drifting: state precisely where you are blocked and ask. "I've worked the example, I can do the single-room version, and I'm stuck on how to avoid re-examining entries I have already handled — is there an angle there I'm missing?" That is a strong sentence, not a weak one. It reports three completed attempts and localises the gap, which usually earns a small targeted hint instead of a large corrective one. ## Rehearsal is the whole point None of this is hard in the abstract; all of it is hard at minute eighteen with an audience. That is why the ladder must be practised until it is automatic — deliberately, on problems you are already stuck on, saying the rung names out loud even when alone. Under pressure you do not rise to your intentions; you fall back on your drills, and a drill you have never spoken aloud is not one you will speak aloud then.