AI-native technical interviews

See how engineers actually work with AI tools

One round of your existing loop, run in a real editor and terminal with an AI tool alongside. You see exactly how the candidate directed it, scored against a shared rubric with the evidence attached.

Early accessOnboarding first teams
By approvalEvery workspace reviewed
90 minOne scheduled session
Live session · Senior backend engineer 28:14
1  def reserve(store, ctx, item_id):
2      tx = store.db.begin(ctx)
3      if tx is None: raise ReserveError()
4      # idempotency key guards double reserve
5      row = tx.execute(LOCK_SQL, item_id).one()
6      tx.rollback()
7      return row
Token usage8.2k
Prompts6
Reviewing approach: locking strategy under contention
The problem

Interview signal broke when AI became normal

Every stage in a standard loop now measures something other than the job. Candidates know it, and so do your engineers.

Take-homes measure free time

A great submission and a generated one look the same. You end up interviewing the code, not the engineer who sent it.

Whiteboards ban the daily tools

Banning the assistant tests memory, not judgment. The skill you are actually hiring for, directing a model well, never shows up.

Nobody scores the skill that matters

Engineers direct a model every day, then get hired on a loop that never looks at it. RecruIQ drops into your existing pipeline and scores that work.

Where it fits

One round, in place of the take-home

RecruIQTake-homeLive whiteboard
Reflects the real workflowEditor, terminal, AI toolsLooks the same either wayNo tools allowed
Shows how AI tools were usedEvery prompt recordedInvisibleNot allowed at all
Resistant to ghost-writingFull session recordedNoYes, by exclusion
Comparable across candidatesOne shared rubricReviewer-dependentPanel-dependent
Candidate time30–90 min a stage, scheduled4–8 hoursMultiple rounds
How it works

Three steps from setup to a decision you can explain

01

Create the opening

Pick a task, select the language, add interviewers. Set it up once and reuse it for every role on the team.

02

Candidate works for real

A real editor and terminal, plus an AI tool with a visible token meter. Keystrokes and prompts are recorded as they work.

03

Review the report

A rubric-based report: a score on each of the five axes, how the candidate used the AI tool, and the moment in the session behind every score, so you can check the reasoning rather than take it on trust.

Session record

Watch the process, not just the diff

Play back the whole session. Every prompt and rewrite is marked on the timeline, so your debrief runs on what actually happened rather than what anyone remembers.

Prompt log alongside the code that resulted
Token usage across the whole session
Shareable timestamps for your hiring review
Priya Raman · Technical screen42:08
07:30Asked the model for a locking strategy, rejected the first answer
15:12Rewrote the generated function by hand
21:40Caught a double-reserve race by reading the query plan
29:55Explained the trade-off against optimistic locking
Axis (weight)PriyaSam
Correctness ×0.30
Prompt quality ×0.20
Verification behaviour ×0.20
Architectural judgment ×0.20
Code quality ×0.10
Composite84 / 100vs62 / 100
Rubric reports

Compare candidates on one rubric

Same task, same five axes, same weights, so two candidates are genuinely comparable. RecruIQ’s model reads the session and scores each axis; the weights are fixed server-side, not chosen by the model, so a composite cannot be talked upwards. Every score carries a link to the moment in the session behind it, and the hiring decision stays yours. RecruIQ never makes one.

PythonTypeScriptJavaScript
For candidates

An interview engineers do not resent

One scheduled session instead of a weekend of homework. Candidates see the rubric before they start, and work in a normal editor with the AI tool switched on.

Rubric shared up front, so they know what is scored
30 to 90 minutes a stage, scheduled by them
Feedback summary returned either way
What the candidate sees
Pick your own slotAny time in the next 10 days
The rubric, in advance5 axes, fixed weights, no surprises
The tools we provideEditor, terminal, and an AI tool
Integrations

Planned integrations On the roadmap

None of these are live yet. They are what we are building next so stage changes, reports and scheduling sync both ways instead of moving by CSV. Tell us which one you need first and it moves up the list.

Greenhouse
Lever
Ashby
Workable
GitHub
GitLab
Slack
Google Calendar
Security and compliance
Evidence-linked reporting PlannedEvery observation will carry a pointer to the moment in the session it came from
Retention that runsSession recordings are deleted after six months, and a daily job checks that none are left
SSO and SAML PlannedSign-in today is a one-time email link, with no passwords stored at all
Regional hosting PlannedRegion pinned at workspace creation
Pricing

Two plans, both approved before you start

We approve every workspace by hand while we are in early access. No per-candidate fees, and interviewers and observers are always free.

Starter
For a first pilot round, approval required
Contact us
Early-access pricing
Request access
5 interviews per month One hiring manager Standard rubrics and problems Session recordings kept 6 months, then deleted
FAQ

The questions we get first

Cheating assumes the tool is banned. Here it is part of the task. We record every prompt, and the rubric scores how well the candidate directed, verified and corrected the model. Someone who pastes an answer they cannot explain scores badly on prompt quality and verification behaviour, and because each score links to the transcript, you can see why rather than take the number on faith.

Add this round to your loop

Request access and we will set up your first round with you. Or take a 30-minute walkthrough with an engineer first.