Practise the reasoning interviews actually expose

Technical interviews ask you to produce reasoning while somebody watches. It is not enough to recognize a two-pointer solution after seeing it or to understand a dynamic-programming recurrence once another person has written it down. You need to identify the state, explain why it is sufficient, test an edge case, and recover when the first approach fails.

A practice routine built around walkthroughs can create a misleading sense of speed. Nudge changes the default stuck response from opening the full solution to answering one smaller question. The goal is not artificial difficulty. It is to keep enough of the reasoning on your side of the table that the completed problem represents something you could explain again.

• Pass one — solve: state the approach before coding and ask for the smallest useful hint only when the reasoning stops.
• Pass two — explain: after acceptance, describe correctness and complexity without reading the final code line by line.
• Pass three — retrieve: revisit the same pattern later and reconstruct the key invariant from a blank editor.
• Track not only whether you solved the problem, but where assistance entered the process and whether the same block returns.
• Pair extension practice with timed human mock interviews so communication pressure is represented too.

At the planning stage, the useful intervention is rarely code. It is a question about the input shape, a brute-force baseline, or the information the state must preserve. When a learner is stuck after writing code, the useful intervention becomes more local: an invariant, an uninitialized value, or a counterexample. Near completion, one overlooked boundary may matter more than another algorithm explanation.

Nudge's stage-aware model is designed around those differences. The coaching is not a substitute for your explanation; it gives you the next object to reason about. The side panel is useful for talking through a plan, while an inline prompt can keep attention near the line or condition that matters. Pro users deliberately include editor context through @code when that context is needed.

A useful week mixes new problems, retrieval, and communication. Solve a small number of new problems with careful notes about assistance. Revisit older problems without opening the previous conversation. Then explain one solution aloud as if an interviewer has asked why it works. The number of completed problems matters less than whether the reasoning survives the gap between attempts.

Nudge's profile can make parts of that behavior visible through solve history, hint usage, streaks, difficulty, time, and recent activity. It is not a complete interview-prep curriculum or spaced-repetition planner. Keep a simple external review queue if you need one, and use mock interviews to test the communication layer that a solo browser extension cannot reproduce.

• Use a course or textbook when the missing piece is foundational knowledge.
• Use a human mock interviewer when you need feedback on communication, pacing, and collaboration.
• Use a broad AI assistant after the attempt for deeper concept comparison or study-plan generation.
• Use company-specific research when preparing for a known interview format.
• Use Nudge when the immediate job is preserving momentum and reasoning inside a LeetCode attempt.