Mercor publishes the signals that raise your odds. Almost nobody applying has read them. Every item here is a fixable thing, most in under ten minutes.
The published list
From Mercor's own referrer documentation, top candidates typically have:
A complete Mercor profile and a strong résumé
A professional, up-to-date LinkedIn with clear credentials
A profile photo - and a GitHub link if it's relevant to your field
Camera enabled during the AI interview, which is required for verification
No VPN usage during the application process
Their stated reasoning: the more data points you provide, the more confident Mercor is when extending an offer.
The one that surprises people
Turn your VPN off. Running one during the application process counts against you. If you use a VPN by default - a lot of technical people do - this is worth checking before you apply, and it's invisible as a problem otherwise.
Interview Formats by Platform: What to Expect
Candidates on Reddit and LinkedIn frequently report confusion over interview styles. Here is the exact breakdown of the three primary evaluation formats across Mercor and micro1.
🌐
1. micro1 "Zara" AI Interview
Structure: A 20-minute continuous interactive interview with "Zara", micro1's virtual AI interviewer.
Question flow: 1 resume walk-through, 2 technical problem-solving scenarios, and 1 client communication question.
The Zara cutoff trap: If you stop speaking for 4+ seconds, Zara assumes your answer is complete and cuts in with the next question. Use our hold-the-floor phrase: "Good question, let me organize my thoughts..."
Post-Interview: You enter HMR (Hiring Manager Review) where human project managers match your score to client budgets.
⚡
2. Mercor Asynchronous 3-Minute Video Screen
Structure: A 3-minute camera-on recorded response to a specific technical prompt.
Camera verification: Mercor requires a verified webcam feed. Using a VPN will flag your submission for anti-fraud review.
Scoring formula: High weighting on first-person ownership ("I architected" vs "we worked on") and quantified outcomes (% improvements, latency reductions).
Target duration: Aim for 60 to 90 seconds per prompt. Do not rush to fill all 3 minutes if your answer is complete.
💻
3. Coding Take-Homes & Domain Setup Checklist
Before launching technical sandbox assessments:
Environment: Use modern Chrome or Edge with browser extensions disabled to prevent sandbox collision errors.
Screen monitoring: Full-screen recording is active; closing the tab or switching focus can terminate your evaluation session.
Prompt red-teaming: For RLHF roles, you will be evaluated on your ability to spot hallucinations and enforce JSON schemas with few-shot prompts.
Final Readiness Quiz
Before you apply, let's verify you absorbed the core principles.
1. What is the "Headline First" rule?
2. During the AI Interview, you should:
3. When writing a prompt for an LLM to output JSON, you must:
✅
You are ready.
You understand the principles of high-end RLHF work.