Two example sessions, shown the way a finished interview is: the problem, the conversation, what was submitted, the tests or the whiteboard, and the feedback. This page is stored example data written for it; opening it calls no AI model, and the interviewer's wording in a real session will differ.
Every interview allows 30 questions, 3 hints and 3 graded submissions per problem. See every limit by plan or compare plans.
The problem workspace
What practice looks like on a free problem: pick a language, take a hint, run the code. A failing case shows the input, what was expected and where your output differs; a case that raised an exception is marked as a runtime error instead, with the exception.
Given an array of integers nums and an integer target, return the indices of the two numbers that add up to target. Exactly one such pair exists, and you may not use the same element twice. Return the two indices in increasing order.
Input is [nums, target], e.g. [[2,7,11,15],9].
11 languages in the editor; the starter code changes with the language.
class Solution:
def twoSum(self, nums, target):
# Return the two indices (increasing order) that sum to target.
pass
Hints, one level at a time
Hint 1
A brute-force double loop is O(n²). Can you trade memory for speed?
Hint 2
While scanning, store each value's index in a hash map. For the current value x, check whether target − x has already been seen.
Three runs of the same problem
Recorded, not live: each result is what this code produced through the judge's own test harness with Python 3.11 on 2026-11-20. A live run on the judge uses Python 3.14 and needs an account.
Stores the value before checking for its complement, so a number can pair with itself.
class Solution:
def twoSum(self, nums, target):
seen = {}
for i, x in enumerate(nums):
seen[x] = i
if target - x in seen:
return [seen[target - x], i]
1/3 Passed
Execution summary
Result (1/3)Failed
Total
3
Passed
1
Failed
2
Wrong AnswerYour code ran to the end and returned a different answer.
Input:
[[3,2,4],6]
Expected Output:
[1,2]
Your Output:
[0,0]
Difference:
Expected[1,2]
Got[0,0]
Sample mock interviews
Two sample sessions in the format a finished interview is saved in. They are written for this page, not recordings of a candidate; the scores and the board outline come from the same scoring code a real session uses.
Read one line of lowercase letters from standard input and print the length of the longest substring in which no character repeats.
Example: abcabcbb → 3 ("abc"). Example: bbbbb → 1.
Constraints: 1 ≤ length ≤ 100,000.
Interviewer
Welcome. Take a minute with the statement, and think out loud as you go.
Candidate
Can the line be empty, or contain anything other than lowercase letters?
Interviewer
It has at least one character, and only a to z.
Candidate · Hint used
Could I have a hint?
Interviewer · Hint used
Hint: when you see a repeat, how far back must the start of the window move?
Candidate
Past the previous position of that character, so I keep the last index of each letter. O(n) time, O(1) space for 26 letters.
Submitted code (python)
import sys
s = sys.stdin.readline().strip()
last = {}
start = best = 0
for i, ch in enumerate(s):
if ch in last and last[ch] >= start:
start = last[ch] + 1
last[ch] = i
best = max(best, i - start + 1)
print(best)
Tests on the judge: 4 of 4 passed (abcabcbb, bbbbb, pwwkew, 100,000 characters).
84/100 · Meets the bar
Strengths
Asked about the input before coding.
Sliding window with the last index of each character: linear time, constant extra space.
Stated the complexity unprompted.
To improve
Walk through one example by hand before running the tests.
One hint was used to find how far the window start moves.
Follow-up question: How would you change it if the input could be any Unicode text, not just a to z?
Design a service that turns long URLs into short codes and redirects visitors. 100 million new links a month, 10 billion redirects a month. Codes never change once issued. Discuss storage, caching and how codes are generated.
Interviewer
Welcome. Start wherever you like; I can see your whiteboard as you draw.
Candidate
Reads dominate, so I put a cache in front of the database. Is a redirect allowed to be slightly stale?
Interviewer
Codes never change, so a cached redirect is never wrong. What happens on a miss in your "Redis cache"?
Candidate
The API servers read Postgres by code and fill the cache.
Interviewer
Your "ID generator" sits beside the "API servers". What happens if it is down?
The whiteboard
Loading the whiteboard…
Notes beside the board
Read heavy, about 100 reads per write. Short codes are base62 of a 64-bit id from the ID generator. Reads go to Redis first (TTL 24 h), then Postgres keyed by code; writes go to Postgres, then fill the cache. Shard Postgres by code.