Coding interview report — anonymized completed interview

verdict

Candidate did not advance on LRU Cache. Core get/put behavior and a capacity guard appeared only late , while the optimal hash-map + DLL approach was never independently completed . Some list mechanics were verbalized , but the implementation stayed incorrect: moveToHead repair unfinished and put/eviction/map sync wrong at the end . After the runner failed, no dry-run replaced execution ; get recency stayed verbal only . Timer expired over budget with integrity review for unrelated pastes and admitted design blockage—do not advance.

rubric breakdown

approach

1/5

Score 1: optimal hash-map+DLL not independently demonstrated; stuck on non-O(1) eviction ideas; interviewer help still not internalized.

testing

1/5

Score 1: run infra failed; no substitute dry-run of sample or edges; verification ineffective.

Written follow-ups

3.4/5

2 of 5 written follow-ups correct

LRU Cache

gaps remainmediumhash-maplinked-listdesign

35m 09s

solve time

Coding execution 2 probesCandidate never reached a working or optimal LRU. After restating the prompt they proposed timestamps plus full scans or a sorted map; the interviewer repeatedly volunteered the hash-map + doubly-linked-list design, Node shape, move-to-front, and tail eviction without being asked. Even with that scaffolding, put remained a stub, moveToHead was incomplete/buggy and placed outside the class, and no testing occurred. Unrelated binary-search paste dumps further undermine authorship of the final fragments. Performance sits in the lowest band: little meaningful independent progress.gaps remain

deduction — approach · −0

interviewer volunteered unsolicited help

optimal — Candidate independently derives hash map + doubly linked list with move-to-front and tail eviction for O(1) get/put

deduction — approach · −0

interviewer volunteered unsolicited help; hashmap+DLL, Node, and move-to-front/evict steps were not independently demonstrated

optimal — Unprompted identification of hash map keyed to DLL nodes, O(1) splice to head, O(1) tail eviction

deduction — code_quality · −1

put left empty; moveToHead incomplete, buggy, and declared outside the class; missing imports; head/tail wiring incorrect

optimal — Complete put (insert/update + capacity eviction), correct bidirectional pointer updates, helpers inside LRUCache, dummy head/tail or careful null handling

deduction — testing · −1

no dry-run, edge-case walk, or execution of the sample input

optimal — Trace capacity-2 sample (put/get/evict), empty cache, update-existing, single-element, and capacity-1 cases

proctoring & anomalies

Candidate identity and raw integrity evidence are omitted from this public excerpt.

Private evidence omitted

strengths

+ Eventually stated core get/put API behavior and an invalid-capacity guard

+ Verbalized some linked-list unlink/tail mechanics mid-session

+ Produced partial follow-up discussion of recency updates on get

growth areas

− approach — Independently derive and state hash map keyed to DLL nodes with O(1) move-to-head and tail eviction before coding; do not stall on timestamp/sorted-map scans.

− code_quality — Finish put (insert/update + capacity eviction), correct bidirectional pointer wiring, keep helpers inside the class, and keep map and list synchronized on every mutation.

− testing — When the runner fails, immediately dry-run capacity-2 sample plus empty, update-existing, capacity-1, and single-node cases on the board.

− problem_understanding — Up-front systematic clarification of capacity edges, update-vs-insert, and eviction order rather than late partial restatements.

− integrity/process — Stay on the assigned design; avoid pasting unrelated binary-search prose when stuck.

Anonymous candidate

LRU Cache