Just wrapped the Replit technical screen last week (mid-2026 cycle). Sharing format and difficulty because this comes up constantly and the info out there is outdated.
Format: Two coding problems, 75 minutes total, done in their own Replit IDE inside the browser. No Leetcode or HackerRank. You write actual runnable code. I could use the language of my choice. I picked TypeScript.
Difficulty: Both problems were medium by Leetcode standards. One was a string manipulation problem with a non-obvious edge case around Unicode handling (fitting given what they build). The second was more of a graph traversal thing, probably BFS. Nothing that would be classified Leetcode hard.
What made it slightly tricky was the emphasis on actually running your code and testing it. They're not just looking at your logic in a vacuum. You're expected to write tests or at least manually test your solution. That caught me off guard a bit because I've been doing Leetcode where you just write the function body and hit submit.
Interviewer involvement: One person was on screen with me the whole time, but they mostly let me drive. Asked a clarifying question after problem 1 about time complexity. Didn't hint or steer much.
One note: the coding environment is obviously Replit itself, which feels a little meta. No autocomplete by default, which annoyed me more than I expected. Worth doing a practice run in a vanilla browser editor so you're not reaching for tab-complete muscle memory.
Overall I thought the format was cleaner than most. The problems tested real thinking, not pattern-matching on canned templates.