Meta New Grad Software Engineer Interview Experience — A Proctored OA Where 35 of 40 Questions Were About AI

VO on CodeSignal, camera on, screen shared.

I had AI clean this up a bit — why has the polishing gotten so hard to use lately... it's not that it was hard, it's that the questions were long, with video explanations, so it took time just to understand them.

New grad. HR reached out directly, and they want this filled very urgently — I need to get pushed to the final round within two weeks, and this batch needs to onboard by June at the latest.

For the work assessment, I feel like Meta has gone completely overboard with AI. Out of 40 questions, 35 of them were just asking how you use AI.

1. Tetris Simulator — Problem: in an n×m grid, place shapes A/B/C/D/E in order. Using "top-to-bottom, left-to-right" priority, find the first position where the shape fits — no overlapping, no rotation. Core approach: brute-force simulation + coordinate offsets.

2. Longest Continuous Houses — Problem: on a number line, build houses in the order given by queries; after each build, find the current longest run of continuous houses. The coordinate range is large, but the number of builds is limited. Core approach: Union-Find.

3. Rating Calculator — Problem: starting rating is 1500. Based on the changes in the diffs array, find the highest rating reached during the process and the final rating. Core approach: one pass.

4. Array Reduction — Problem: loop over the array. Each time, find the leftmost nonzero number x, and starting from it, subtract x from each number to the right consecutively until you hit a number smaller than x. Each time, add x to the result. Repeat until the array is all zeros. Core approach: pure simulation (loop simulation).

I ended up just taking a screenshot of it.