Just got through the Palantir coding rounds (SWE new grad pipeline, spring 2026). Sharing everything since the info online is weirdly sparse compared to Meta/Google.
Format: Two separate coding rounds, each 45 minutes, with different interviewers. Not a take-home OA. Both were live, in HackerRank or a shared coding env -- I forget which exactly, it was CoderPad actually.
Difficulty: I'd call it medium-hard LC on average. I got one graph traversal problem (BFS variant, not a trick, just needed clean implementation and right time complexity) and one string manipulation problem that had a recursive structure to it. Nothing that required knowing some obscure algorithm. The hard part was implementing cleanly under pressure while explaining your reasoning.
Both interviewers asked me to walk through my approach before coding. They stopped me once when I started coding too fast and said "tell me what you're thinking first." That's a hint: narrate, don't just type.
What seemed to matter: Could you arrive at a correct solution, not just a brute force Did you handle edge cases (they asked about them explicitly) Could you discuss time/space complexity clearly Did you write readable code, not competitive-programming-style one-liners
I've done a lot of LC hards to prep for FAANG loops. Honestly that over-prepared me a bit. I spent way less time on Palantir's actual problem and more time panicking about what trick I was missing. There wasn't one. Just solve it cleanly.
One thing that surprised me: the interviewer asked me after to refactor part of my solution for readability, not performance. That's unusual. felt more like a code review than a typical coding interview.
I did not move on (got ghosted after round 2, classic). But I felt like the rounds themselves were fair. If anyone has more recent data from summer 2026 loops I'd be curious if the format changed.