Confirm your candidates understand the code they shipped
Sentinel adds a short voice interview right after the take-home, where your candidates walk through their own code on screen and explain the choices they made, so the first review starts with real signal instead of a guess. No webcam, no lockdown, just a conversation about the work.
A quick chat about your work
A short voice conversation about what you built and the choices you made. Just talk it through, there are no trick questions.
We will ask for microphone access in a moment. To answer each question, hold the button or the space bar. It only takes a couple of minutes.
Why Sentinel exists
We build CodeSubmit with coding agents every day, and we want the engineers we hire to use them too. The part that matters is whether the person can stand behind what they shipped, no matter how they built it.
An agent can produce a clean, passing take-home in a few minutes, and working is not the same as understood. Understood is the part you are paying a senior engineer for, and it is the one thing a green test run can never show you.
Did they actually understand it?
The part that is hardest to fake
Their own code on screen, and the reasoning behind it
A real author moves through their code without pausing. Someone who pasted an agent and never looked back tends to stall on the first honest follow-up. Sentinel puts your candidate's own code in front of them and asks what they were trying to do and why they took the approach they did.
- Shows your candidate’s real submitted code, not a generic snippet.
- Asks what they were going for, and why they took the approach they did.
- Goes deeper when the answers hold up, and eases off when they do not.
“Walk me through why you memoized visibleRows. What would force it to recompute?”
On the candidate's side
A check that respects their time
Sentinel runs asynchronously, the moment a candidate submits. There is no extra slot to schedule and nothing to install, just a few minutes of talking through the work they already did.
Sentinel reads your project
A repo-reading agent studies what you just submitted and prepares a few questions grounded in your own code. It takes up to about ten minutes.
A friendly face is waiting
An auto-start countdown begins on your schedule, with the interviewer ready to go. No calendar invite, no extra slot to book.
Just talk it through
Hold the button or the space bar and answer out loud. Your own files appear on screen, and the line in question is highlighted as you go.
About five minutes, on their schedule, capped at 30 turns and 300 seconds so it never drags.
The wrong answer
We did not build a surveillance tool
The reflex in hiring is to watch the person harder. We think that measures the wrong thing, and tells your candidate the wrong thing about how you work.
- Webcam proctoring
Engineers never build under a webcam, so it proves nothing about the job.
- Lockdown browser
It blocks the docs and tools real work depends on, then calls that a fair test.
- Keystroke logging
It measures typing, not whether the person understands what they shipped.
- Screen recording
Hours of footage nobody watches, and still no answer to the only question that matters.
Lock the browser, watch the webcam, log the keystrokes. None of it tells you whether the person can stand behind what they shipped, and all of it tells your strongest candidates that you do not trust them.
Nobody works under a microscope once they have the job, so we would rather ask the thing a decent teammate asks in the hallway. Can you walk me through this?
No webcam, no screen recording, no keystroke logging, no lockdown browser, just a conversation about the code.
So we built the opposite of surveillance
“We love AI coding tools and we want engineers to use them. We just don’t want anyone blindly shipping code an agent produced without understanding it, and a five-minute conversation about the code confirms they do.”
Dominic PhillipsFounder, CodeSubmitWhat lands in your dashboard
A score and a story you can act on
Each conversation returns a 0 to 100 authorship and understanding score plus a short written summary, so your first review starts with evidence instead of a hunch. It sits right beside Warden, and the final call always belongs to your team.
- One scannable score, with the reasoning behind it.
- A short written summary of the strengths and the red flags.
- Extra context for your team, never an automatic pass or fail.
Explained their own code without hesitating. A signal for your reviewers, never an automatic decision.
Add a quiet minute
Hear how your next candidate explains the code they shipped.
Turn Sentinel on for your next take-home and start the first review with a real sense of who is behind the work.
Off until you turn it on, per assignment. The final call always belongs to your team.