A heads-up on AI mistakes.
Get private alerts about possible AI mistakes and known pitfalls. Inspect the evidence and decide what deserves a follow-up.
A new way to interview. A new way to prepare.
Make your
next interview
feel familiar.
Practice AI-assisted interviews in real codebases. Learn to prompt, question, and verify — with a teammate beside you.
Every problem free. One graded assessment per month.
01 Prompt
02 Review
03 Verify
Confidence is built in the loop.
Real questions.
From companies you know.
Original practice exercises adapted from real interview reports.
- Meta
- Amazon
- Stripe
- Coinbase
- DoorDash
- Shopify
- and more
AI can write the code.
You still own the thinking.
AI-assisted interviews ask you to do more than find an answer. Can you make a useful request, spot a shaky assumption, and prove a change works?
nrml gives you a place to practice that entire conversation.
Practice the
whole performance.
Start with a real repository.
Build the habits you'll bring to the interview.
A small bug. The whole workflow.
Open a file. Ask a question. Review a change, or write your own. Then put it to the test.
src/path_search.tsEditable function excerpt
Changes stay in this demoTypeScript · Spaces: 2
Run the starter first. Review a sample AI change, then run the same checks again.
- Find an interior start
expected 5 - Find the start on the top row
expected 0 - Find the finish on the top row
expected 3 - Report a missing marker
expected -1
Review the proposed changes
src/path_search.ts · Sample AI proposal
Start the scan at the first cell, so top-row markers are included.
Current code
Proposed code
The editor changes only after you apply this proposal.
Interactive example · Sample AI replies; checks run on your edits.Try a full practice session
On the other side of the interview
Better questions.
More useful feedback.
Be the interviewer you’d want to practice with. Get a second perspective on the AI, capture what you notice, and ask your own questions in private.
Practice with someoneAn AI copilot just for you.
Ask about the candidate’s approach or find a sharper follow-up question. Your copilot conversation stays hidden from the candidate.
Feedback while it’s fresh.
Capture observations as the conversation unfolds. Keep notes private, then share your chosen feedback after the interview.
Screenshots from nrml with example session data and scripted AI responses. Alerts surface concerns to investigate.
After the sessionKnow what to
Know what to
practice next.
Get feedback on the work and the way you got there: your tests, your prompting, and your engineering judgment.
Explore the sample reportAn illustrative report with example feedback. It is separate from the interactive examples above.
Free includes one graded assessment per month. Pro adds unlimited grading, full explanations, solution walkthroughs, and session replay. Compare plans
Further reading
Browse all articles Latest articles
Worked examples for AI coding, debugging, and interview preparation.
AI interview practice: check that generated tests can fail
Audit AI-generated JavaScript tests for empty-result blind spots, shared assumptions, and missing guarantees using literal examples and deliberate mutations.
Read articleAI debugging: test a local day across daylight saving
Fix a Python report that treats every local day as 24 hours. Verify UTC boundaries, repeated hours, exclusive endpoints, and rejected naive timestamps.
Read articleAI coding: keep repository text from changing the task
Work through an injected repository instruction, constrain a proposed patch to approved files, and verify behavior without treating source text as authority.
Read article
Before you
jump in.
What can I do for free?
Practice every problem, run code and tests, and get one graded assessment per month. Pro includes unlimited graded assessments, full explanations after an attempt, solution walkthroughs, and session replay. AI usage has a separate allowance.
What is an AI-assisted interview?
It's an interview format where you work with AI while solving an engineering task. You still need to understand the problem, guide the tools, review changes, and show that the result works. The exact format depends on the interviewer.
What does practice look like?
You start in a real repository with a problem to work through. Use the integrated AI in Ask or Edit mode, inspect its changes, and run commands and tests in the terminal.
Can I practice with someone else?
Yes. Guests can join without an account, edit code and diagrams, contribute shared notes, and follow another participant. You can also keep private notes for your own preparation.
What feedback do I get?
Reports look at test outcomes, prompting, and engineering judgment. Session replay is available with Pro so you can revisit the decisions you made.
Meet the format.
Find your footing.
Your next interview starts with a little practice.