I saw someone tell this story recently, so I am going by their telling of it.
They were in an interview. The interviewer asked: "do you use AI?"
They took it badly. The way they told it, the question itself was out of
line - an attempt to catch them out. So they answered with a question of
their own: are you going to use a stove instead of a microwave? The world
moves forward. They will stop making stoves eventually.
They told the story as a win. I do not think it was one, and it took me a
while to work out why. It is not because using AI is wrong, or because
answering back in an interview is wrong. It is that a specific and very
common mistake is happening in that exchange, and I have made it myself.
They answered a question nobody asked
"Do you use AI?" is a question about method. It has a boring, factual
answer: yes, here, for these things, and I check it like this.
What they heard was "are you a real engineer?" That is a question about
status, and status questions feel like they need to be won rather than
answered.
The gap between those two readings is where the whole thing goes wrong. The
interviewer asked about a tool. The candidate defended an identity. And once
you are defending an identity, the actual content of your answer stops
mattering to you - which is a shame, because it still matters to the person
across the table.
Why the misread is reasonable
It would be easy to say they were being oversensitive. I do not think that
is fair, because the question genuinely has a charge on it now.
"Do you use AI" in 2022 was small talk. Today it is a question people ask to
find out which side you are on, and it gets asked from both directions.
Somewhere there is an interviewer who wants to hear "no, I write everything
myself" and somewhere there is another who wants to hear "yes, constantly,
I ship three times faster". Both of those interviewers exist and neither of
them is signposted.
So the candidate is not paranoid. They are guessing which room they are
standing in, under time pressure, with a job on the line. That is a real
problem and it is worth naming rather than mocking.
But guessing wrong has a cost, and the cost here was the whole answer.
The analogy argues against him
This is the part I keep coming back to.
Stove versus microwave is not old versus new. They are not the same
appliance. A microwave reheats things quickly and unevenly. A stove sears,
reduces, browns, holds a temperature. Every professional kitchen has both,
and nobody in one thinks the microwave is winning.
So the analogy, chosen to prove that the interviewer was behind the times,
quietly concedes the interviewer's point: what you actually want in a kitchen
is somebody who knows which one this dish needs. That is precisely what the
question was trying to find out.
An analogy that feels like it ends the argument is worth checking twice. This
one ends it in favour of the other person.
"The world moves forward" is a way of not answering
The second half - "they will stop making stoves" - is an inevitability
argument, and inevitability arguments have a particular smell in interviews.
The question was about your practice. The answer was about history's
direction. Those are not the same subject, and swapping one for the other
is one of the oldest ways to avoid saying something specific.
It also cannot be wrong, which is the problem. "Things will keep changing" is
true in every possible world, so it tells the listener nothing about you. An
interviewer is not collecting predictions. They are trying to work out what
you will be like on a Tuesday when something breaks.
What the boring answer would have sounded like
Something like: "yes, daily. It writes most of the first draft. I read all of
it, I am careful with anything touching auth or money, and I have been burned
by it inventing a library function that did not exist, so I check the imports
first now."
That is not an impressive answer. It is specific, it admits a scar, and it
shows a line. It would have taken twenty seconds and it answers the question
that was actually asked.
The instinct that specificity is weakness gets a lot of good engineers in
trouble. In an interview, specificity is the entire product.
Who can afford to fire back
One more thing, and it is uncomfortable.
Answering an interviewer with a rhetorical question is a status move, and
status moves are priced. If you are senior, scarce, and the company needs you
more than you need them, it reads as confidence and it often works. If you
are not, the same words read as brittle.
Nothing about that is fair. It is just how the room works, and it is worth
knowing the price before you spend it.
Where the line actually is
The line has quietly moved, and it is no longer usage.
Nobody serious is being judged for using AI in 2026. What people are being
judged on is where they put the check: what they read, what they refuse to
ship without understanding, and whether they can tell when the output is
confidently wrong.
That is a much harder thing to interview for than yes or no, which is
probably why interviewers keep reaching for the crude version of the
question. If you get the crude version, answer the real one anyway. You will
be the only candidate that day who does. :)
Top comments (2)
Your reading of the analogy is the part I keep turning over: it was chosen to end the argument, and it actually hands the point to the interviewer. Two tools that do different jobs is an argument about judgement, and judgement was exactly what the question was asking for.
The addition I'd make is that a good answer here needs one falsifiable piece in it. Not "I review everything" — at any real volume that is not checkable, and interviewers can smell it. The answer I respect most is a named task from last week where the tool stayed in the box and why it would have made that one worse. It is also an asymmetric question: "no, I don't" now reads as a missing skill and "yes, constantly" reads as outsourced judgement, so the candidate is answering under a prior either way. Guessing the room is hard; naming the boundary out loud is the move that works in both rooms.
Agreed, and worse than uncheckable - at volume "I review everything" is probably untrue, and they know it, so it costs you. The asymmetry point is the one I missed. Though once "when didn't you use it" is a normal question, won't people just prepare one, like "greatest weakness"?