Wordle-Style 5-Letter Minimum-Guess Strategy

Phone screen coding: given a black-box oracle that tells you whether a specific letter is at a specific position in a hidden 5-letter word, design an algorithm to identify the word in as few oracle calls as possible. Follow-up assumes a dictionary of candidate words is available. Requirements Hidden word: a 5-letter string from a..z. Oracle: query(pos: int, letter: char) -> bool returns whether position pos (0-indexed) holds letter. Return the hidden word. Primary metric: minimize the number of query calls. Brute-force baseline: 5 × 26 = 130 calls. Follow-up A dictionary of candidate 5-letter words is available; design the strategy to use bigram/letter-frequency information to cut the expected number of calls. Examples HiddenWord = "GRAPE" query(0, 'A') → false query(0, 'G') → true // first letter found ... Notes Without the dictionary, query positions in order; per position, query letters in descending frequency (e, t, a, o, i, n, ...) and stop as soon as one returns true. With the dictionary, maintain a candidate set; at each step pick the (position, letter) query whose true/false outcome splits the candidate set most evenly (entropy maximization). Mention information theory: each query yields ≤ 1 bit, so the absolute lower bound on calls is ceil(log2(|candidates|)). Don't waste minutes on a closed-form optimum — the interviewer mostly wants you to articulate the entropy heuristic and pick one reasonable scoring function. Preparation Drill information-theoretic puzzles (egg drop, twenty questions) to internalize the entropy-split heuristic. Practice maintaining a candidate set under positive/negative oracle feedback (filter on each query result). Be ready with a one-liner on Wordle's known optimal first guess (SOARE / CRANE) for intuition; do not hard-code it.