Interviews are getting harder to read, because AI means almost anyone can produce working code. According to TechCrunch, HackerRank's AI interviewer, called Chakra, has run more than 500,000 interviews in beta. If you hire developers, the practical lesson is simple: look at a supervised assessment score first, then use your call time on judgment.
This article explains what HackerRank announced, why it matters, and what to do in your next hiring round.
What did HackerRank announce?
HackerRank made Chakra generally available to customers on Monday after about six months in beta. Chakra is an AI agent that runs the interview. It watches the candidate work and judges how they got to an answer, not only the answer itself.
The candidate gets a task in a real code repository. They work in a canvas that includes an AI assistant. Chakra asks follow-up questions, such as why they chose one approach, or what changes if a new constraint appears.
According to the report, Snowflake, Snorkel and Capgemini were among the companies that tried it. HackerRank's CEO says it replaces three steps: a recruiter screen, a take-home test and an engineer interview.
Why are interviews harder to read now?
The CEO put it this way in the TechCrunch piece: because of AI, anybody can produce an artifact. A clean solution no longer proves the person understands it.
So the useful signal has moved. It is now about thinking: how the person frames the problem, checks the output and changes direction when something breaks.
That is our view as well. A right answer is cheap. The reasoning behind it is what you are hiring.
What should you ask instead of quiz questions?
Quiz questions test recall, and recall is easy to fake or look up. Questions about real decisions are much harder to fake.
Try these in the first five minutes of a call:
- Why did you choose that approach over the other options?
- What would you change if you built it again?
- What broke the last time you shipped something like this?
- How would your solution change if the traffic doubled, or the deadline halved?
Listen for specifics: names of tools, trade-offs, a mistake and what they learned. Vague answers are the warning sign.
Can an AI interviewer replace the human call?
Not according to HackerRank. The CEO told TechCrunch that Chakra scores candidates and humans make the final decision. The human call is for deciding whether you want to work with the person, and for answering their questions about your team.
There are open questions too. TechCrunch notes that applying the same criteria every time does not make a system free of bias, and that regulators are looking at automated hiring tools. New York City, for example, requires an independent bias audit and notice to candidates for certain tools.
If you use any automated scoring, check the rules where you and your candidates are based. This is a point to raise with your own advisers, not something we can settle here.
How to use a score before you book the call
A score is only useful if you read it before the conversation, not after. Here is the order we suggest.
- Technical score first. Check it was taken under supervision and was specific to the role. A generic quiz tells you little.
- Communication score second. A strong engineer who cannot explain a trade-off will slow a remote team down.
- Video introduction. Two minutes tells you whether you can follow their explanations.
- Verified history. Confirm education and work history are checked, not self-reported.
- Then the call. Spend it on the why, the what-would-you-change and the what-broke questions above.
Every visible profile on Talent Navigator has cleared all four: a role-specific technical test under camera supervision, a separate communication score, a video introduction and verified education and work history. For more on testing and interviews for specialist roles, see our guide to hiring machine learning engineers and our key stats on AI developer interview questions.
What to do this week
- Write down the three questions you ask in the first five minutes. Replace any that a search engine could answer.
- Add one question about a real failure the candidate has handled.
- Decide your pass mark for technical and communication scores before you see any profiles.
- Stop running a separate screening call for people who have not yet shown a score.
- Check whether any automated tool you use needs candidate notice or an audit in your location.
This keeps your calls short and your decisions consistent.
Frequently asked questions
Does letting candidates use AI make cheating easier?
HackerRank says the opposite showed up in its data. Its CEO told TechCrunch that suspicious-activity flags were 70% to 80% lower in Chakra interviews than in comparable traditional assessments, though results varied by geography and seniority. That is HackerRank's own finding, so treat it as one data point.
Is a high technical score enough to hire someone?
No. A score tells you the person can do the work under supervision. You still need to judge communication, how they handle trade-offs and whether they fit your team, which is why we show a communication score next to the technical one.
Does Talent Navigator use an AI interviewer?
No. We use a supervised, role-specific technical test, a separate communication score, a video introduction and verified history. You search, shortlist and interview candidates yourself, with no recruiter in the middle.
If you have an open role, you can Browse vetted talent and read the scores before you book a single call.