just did the NYT online assessment as the first technical gate for a software engineer role. wanted to document it because the information i found online was pretty scattered and some of it seemed outdated.
format: HackerRank-based OA. timed. two problems, 90 minutes total. you can use any language.
problem 1: a string parsing problem. given a structured news article format (honestly it felt custom to NYT, or at least themed that way), you had to extract and transform certain fields. medium difficulty, probably closer to LC medium-easy. the tricky part was handling edge cases in the input format, not the algorithm itself.
problem 2: a graph/tree problem. you're given a dependency structure (again, kind of themed around content relationships) and need to do a traversal to find something. genuine LC medium. BFS/DFS, standard stuff, but the edge cases were non-obvious. i almost missed the disconnected-components case.
time: 90 minutes for 2 problems felt comfortable. i finished in about 65 minutes. don't rush if you don't need to.
did it feel like a filter or a bar?: filter. i think they use the OA to cut down volume before scheduling human time. it's not trying to find genius, just verifiable ability to code.
my prep: i was doing 2-3 LC mediums a day for about three weeks before this. mostly graph + string stuff. that was sufficient. i don't think you need to be grinding LC hard to clear this gate.
one thing i noticed: no system design in the OA, no behavioral either. it's purely algorithmic. the system design and behavioral rounds come later in the loop if you pass.
i'm now waiting on recruiter feedback to find out if i move to the phone screen. fingers crossed. will update this thread.