Meta AI-enabled Coding Interview Questions and Guide

Meta has fully rolled out a new format of interview round in its full loop - AI-enabled Coding, where candidates get to work with AI assistants while solving a coding problem.

Who gets the Meta AI-enabled coding interview?
AI-enabled interviews appeared in the full loop for the following candidates
SWE, E4 - E7
SWE, Machine Learning
SWE, Product
MLE
Engineering Manager, M1 - M2

What is the AI-enabled interview like?
Duration - 60 minutes
Environment - CoderPad
Coding languages - Python, Java, C++, C#, Kotlin, TypeScript
3 panels from left to right - file explorer, code editor, rightmost panel with 3 tabs
[problem description], [program output] and [AI assist].


Problem breakdown
0. Intro
Interviewer introduces the interface.

1. Troubleshooting
Run the tests for the first time to locate the bug(s), which usually can be fixed by modifying a couple lines. Common bugs include off-by-1, a wrong operator, swapped conditions, etc.

2. Core solution
Implement the core algorithm to the problem - such as maze searching, scoring computation for a card game and string processing. Some of the optimal solutions can take 100+ lines of code, where AI help is required.

3. Followup cases
At this stage, your solution gets challenged by more test cases, which can affect the choice of the core logic for your solution.
Some tests introduce new assumptions, and you will be expected to rework on the solution to support the additional requirement. Sometimes, this would require implementing an absolutely different strategy for the solution.
Some test cases can be very large, requiring an optimization on your existing solution.

Scoring criteria
1. Pass ALL basic tests. Pass as many test cases at the followup stage as possible. It is alright to miss a big test file. A candidate whose solution stuck with a large test case, was permitted to run the file by part, and still moved forward to team matching.
2. Demonstrate your problem solving skills. Do not just throw everything to AI and copy paste AI's solution. Determine on the strategy of your own, provide reasons, ask AI for the implementation and review it.
3. Prompt for clean and well-structured code.
4. Code review and refinement.
5. Communication. Be responsive, reason for your decisions, talk through the answer provided by AI, keep the interviewer on track with what you are on to.

Challenges
1. In addition to knowing the traditional leetcode-style problem solving, the AI-enabled interview requires experience working with AI and prompt engineering.
2. Reengineering and reviewing the long solution provided by AI within the time constraint, especially on the final stage working with the imported test cases, is the major hinder.
3. 2-way communication increased to 3-way. Multi-tasking among interpreting AI's answer, communicating with the interviewer and conveying your thoughts can be hectic.

...