Meta
Speed and breadth: about two problems in 40 minutes, code not run, a broad spread of standard data structures, and no dynamic programming (Meta says so itself). Since late 2025, one coding round is often a 60-minute AI-enabled round in a multi-file codebase. Coding decides hire or no-hire, design largely sets the level, and failing behavioral is a no-hire.
The loop
Section titled “The loop”| Stage | Format | Evidence |
|---|---|---|
| Online assessment (some candidates, since 2025) | 90-min proctored CodeSignal | thin |
| Tech screen | 45 min, usually two easy/medium problems in CoderPad with execution off | official length; reported format |
| Onsite | “4 to 6 of 45 minute interviews”: coding, design, behavioral | official |
| E4/E5 onsite | 2 coding (one may be AI-enabled), 1 design, 1 behavioral | reported |
| E6 onsite | Adds a leadership assessment and a second design round | reported |
| Team match | Before the offer | reported |
- AI in standard rounds: banned. “Outside assistance includes… using AI tools, like ChatGPT.” No background blur. official
- Language: any. “Scripting languages tend to be easier.” official
Standard coding rounds
Section titled “Standard coding rounds”Straight from Meta’s guide official:
- “You should expect to solve about two problems in the course of about 40 minutes.”
- “The interviewer will not compile your code… Finding and catching bugs in your code is a positive sign.”
- “We do not ask dynamic programming questions, so don’t spend time prepping for that technique.”
- Know arrays, stacks and queues, hash tables, trees and BSTs, graphs, heaps. “Prioritize breadth over depth.”
| Topic area | Study first |
|---|---|
| Arrays, strings, hashing (dominant in reports) | Hash maps, Strings, Prefix sums |
| Two pointers and sliding window | Two pointers, Sliding window |
| Trees and graphs (BFS/DFS) | Binary trees, BFS, DFS |
| Stacks; parsing | Stack, Parsing |
| Intervals and heaps | Intervals, Heaps |
Reported round shape: one more algorithmic problem plus one clean-code problem. Minor null or empty-input bugs can be forgiven, but an incomplete algorithm usually can’t, and failing both problems is usually a no-hire. reported
The AI-enabled coding round
Section titled “The AI-enabled coding round”Piloted from October 2025, with Meta’s recruiting quoted as saying “it’s not ‘cheating’, it’s how the job is evolving.” reported It’s not in the official guide, and sources disagree on who gets it, so ask your recruiter.
- Format: 60 minutes in CoderPad with a file tree, runnable code and tests, and an AI chat panel. The AI can’t edit files; you paste code in yourself. It replaces one of the two coding rounds. reported
- Problem: one multi-part problem in an unfamiliar codebase: (1) fix a bug, (2) implement the core logic (100+ lines, harder than a LeetCode medium), (3) optimise for bigger inputs. Reported families: maze and pathfinding, backtracking search, recommendations. You don’t have to finish every part. reported
- Scoring: the same four focus areas. You “should use AI, but need to show you understand the code”; using it “as a crutch” gets marked down. reported
- How to practise: read unfamiliar multi-file code fast, state your plan before you prompt, and review every line the AI writes. Study Debugging, Backtracking and Shortest paths.
Design and behavioral
Section titled “Design and behavioral”- Two design flavours official:
- System design: “distributed systems, scalability, performance and efficiency” (a URL shortener, chat, worldwide video distribution).
- Product architecture: “API design, usability and utility”: data models, client-server design, accommodating product change (a feed or chat API, an email server).
- Product and full-stack engineers usually get product architecture; infra and backend engineers get system design. reported Study API design and Extensibility.
- Behavioral official: five focus areas: resolving conflict, growing continuously, embracing ambiguity, driving results, communicating effectively. Use STAR with “tangible/quantifiable RESULTS”, and outline 2-3 major projects in advance. At E5+, projects should run longer than a quarter and come with metrics. reported
How you’re scored
Section titled “How you’re scored”Coding has four official focus areas official:
- Communication
- Problem solving: compare solutions, data structures, time and space
- Coding: organised, executable code
- Verification: test cases, an argument for correctness, finding your own bugs
Design is scored on problem navigation, solution design, technical excellence and technical communication. official
Why people fail reported
- Too slow for two problems in 40 minutes.
- Not verifying code that will never run.
- Ignoring the interviewer’s hints.
- AI round: pasting AI output without reviewing it, or getting lost in the codebase.
- Product architecture: vague APIs and data models.
Your plan
Section titled “Your plan”Use the Sandbox’s Meta filter. Speed drill: two problems back to back, Level 1 of each, inside 40 minutes, tracing by hand before you press Run. Skip DP. The Sandbox has no AI round, so practise that on a small open-source repo with an AI chat open but not editing files.
| 2 weeks | 4 weeks | |
|---|---|---|
| Week 1 | Daily pairs: Pair Targets + Bracket Repair, Subarray Sums + Palindrome Repairs, Integer Parsing + Word Abbreviations, Tree Queries + Nested List Weights | Same pairs, one per day, plus Array Transforms + Anagram Groups |
| Week 2 | Calendar Interval Toolkit + Island Survey, Function Time from Logs + Path Resolver, Warehouse Robot Routes + Combinations Toolkit, Expression Calculator. One AI-round drill. Two design sessions. 6 STAR stories across the five focus areas | Graphs and search: Island Survey, Warehouse Robot Routes, Alien Alphabet, Combinations Toolkit |
| Week 3 | Function Time from Logs, Meeting Room Allocation, Merging Sorted Feeds, Randomized Bag, Expression Calculator. Two AI-round drills. Three design sessions | |
| Week 4 | Full mocks: 2 × (two problems in 40 min), one AI drill, a behavioral mock |
Ask your recruiter: Will one of my coding rounds be AI-enabled, and which models and languages? System design or product architecture? Is there an online assessment?