Officially

Meta just rolled out a new AI-enabled coding round that replaces one of the traditional onsite coding rounds (two LeetCode-style problems in 35 mins). Instead of classic meta-tagged algo problems and their variants, from what I've gathered you'll get one of three scenarios: building a feature from scratch, extending an unfamiliar multi-file codebase, or debugging broken code under time pressure. All with AI assist, plus real execution and testing (Python candidates: brush up on unittest if you haven't already). I started seeing members of my interview prep Discord getting this round over the past week or 2, and since there's not much info out there yet, I spoke to them to gather as much insight as possible. Below I discuss what I've gathered so far, you can find practice questions here, hope it helps. ## The Basics * 60-minute CoderPad session with an AI-assist chat window and model switcher (similar to Copilot Chat, but simpler). * Confirmed available in interviews: GPT-4o mini, Claude 3.5 Haiku, Llama 4 Maverick, GPT-5, Gemini 2.5 Pro, Claude 4.5 Haiku, Claude 4 Sonnet, Claude 4.5 Sonnet * One thematic question with multiple checkpoints or stages (so it can be a multi-part question), not two separate LeetCode problems. Empirically clearing 3 of the checkpoints seems to be necessary (not sufficient) to clear this round. I've seen people that cleared 3 checkpoints get called to redo though. * You get a mini multi-file codebase (for Python: multiple .py files plus requirements.txt) * You can run and debug code in real-time. * Started appearing early October 2025 for SWE and ML; likely rolling out to Production Engineers soon ## What This Round Actually Tests > What do we look for? > The AI-Enabled Coding Interview will assess your performance on the following four focus areas: Problem Solving, Code Development and Understanding, Verification and Debugging, and Technical Communication. > > • Problem Solving: Are you able to clarify and refine problem statements? Can you generate solutions to open-ended and quantitative problems? > • Code Development and Understanding: Are you able to navigate a codebase to develop and build on working code structures and to evaluate the quality of produced code? Can you analyze and improve code quality and maintainability? Does code work as intended after it is executed? > • Verification and Debugging: Can you find and mitigate errors to ensure code runs/functions as intended? Are you able to verify solutions meet specified requirements, leveraging test/edge cases and handling errors and exceptions? How well do your unit tests run? > • Technical Communication: How well can you communicate reasoning, discuss technical ideas, ask thoughtful questions, and incorporate feedback? ## What You Need to Know About the AI * It is not a frontier reasoning model. Expect hallucinations, suboptimal suggestions, and missed edge cases * The AI can see all code in your editor (no copy-paste needed) * This is not a prompting test. You are evaluated on problem-solving and verification, not AI expertise * Some candidates barely use it and excel; others use it heavily for boilerplate and also excel * The AI is great for: boilerplate, parsing, scaffolding, heavy typing, and help with debugging and navigating the codebase. * AI struggles with: 100% accuracy, algorithmic optimality, edge cases, and deep reasoning ## Things To Avoid Doing * Letting AI drive and do all the work: for example, pasting large outputs without reviewing them line by line * Skipping tests: eyeballing code instead of actually running it * Giant code dumps: requesting 100+ lines at once that you cannot verify * Long silences: going quiet without indicating to the interviewer that you are taking time to think or to do x, y, z. Keep the interviewer in the loop * Ignoring regressions: only re-running the last failing test instead of the full suite * Nonstop narration: talking through every keystroke ## How to Prepare * Ask your recruiter for the practice CoderPad (it has the AI-assist tab and model switcher) * Practice three scenarios: * Building from scratch * Extending unfamiliar multi-file code * Debugging broken code under pressure * Get good at rigorously laying out edge cases: empty input, large values, duplicates, invalid data * Get familiar with the AI-models beforehand: Know which one you will use for different tasks like (writing tests, debugging etc). * Be ready to write test cases or edit existing test cases in the unit test framework for your language: * Python: unittest * C++: GoogleTest (gtest) * C#: NUnit * Java: JUnit If you've taken this round and have insights that would help the community, please share. Best of luck. Edit: - You can try out practice questions here, see the "AI-enabled coding" section - Added a detailed blog post with guidance on how to prepare - link to interview prep Discord