Compare · Approaches
Capifiq vs. bolt-on detection agents.
A detection agent installs on the candidate's machine and watches it while your interview runs somewhere else: a general-purpose meeting tool, a separate coding environment, a shared screen. It's a real approach and a useful one. But it inherits a shape: it can only see the machine, and everything the interview actually happens in sits outside its view. Compared honestly.
The shape of the approach
What a bolt-on agent can and can't see.
This isn't a knock on any particular product. It's what follows from watching a meeting you don't own.
Bolt-on agent
Watches the machine. Doesn't own the room.
- Sees what runs on the candidate's computer
- Works with whatever meeting tool you already use
- Quick to add without changing how you interview
- The coding round, the screen share and any external editor happen outside it
- It can flag that something is running, showing what it displayed is a different capability
- It has no say in what the candidate is asked, so it can't affect help that never touches the machine
Capifiq
Owns the room, so there's nowhere outside it.
- The interview runs on Capifiq: video, shared code pad and whiteboard in one session
- No separate editor or unmonitored share window to move the hard part into
- Only a physical camera can join, so a synthetic feed can't enter the call
- The interviewer sees every connected display, not just the shared one
- Capifiq writes the questions, so it can act on help that never touches the machine
The structural difference
A layer added from outside stays one layer behind.
A browser plugin can inspect the browser. A bot that joins a call can observe the call. An agent on the machine can see the machine. Each is bounded by where it sits, and the tactics that matter most are chosen precisely because they fall outside one of those boundaries.
Adding a detection layer to a meeting tool you don't control means the interview and the integrity check are two separate things that have to be kept in step. Running the interview on Capifiq removes the gap, because there aren't two things.
The question to ask either way
Three things worth testing before you buy anything.
Including us. These are the questions that separate a demo from a deployment, whichever direction you go.
Catch something it has never seen
Ask for a live test with a renamed build, a recompiled open-source fork, or a private one-off: on a machine that wasn't set up in advance. A tool count tells you what a product can name, not what it can catch.
Show the content, not just the alert
"A hidden tool was running at 10:24" and "here is what it was displaying at 10:24" are different claims, and only one of them survives a candidate disputing the finding.
Answer the phone question
Ask what happens when a candidate props a phone against the monitor and reads answers off it. Nothing on the machine changes. Any honest answer starts by admitting no software sees that.
Our answer to the third one: we don't claim to see the phone. We changed what it's useful for: the Question Engine builds the interview questions so a snapshot of the screen doesn't produce an answer, and the follow-up the interviewer asks next was never on screen at all.
Get started
Run the test on your own interviews.
Bring the three questions above. We'd rather you asked them.