Interviewing off leetcode you already memorized isn't cheating, it's the job

Someone was accused of cheating because they were able to solve a difficult problem very quickly. Not for playing with AI. Not for copying from a screen. Just for excelling at it.

The theater everyone agreed to perform

There's a survival strategy floating around interview forums that broke my brain a little. You see the issue right away. How do you react to that? You act as if you don't. One applicant nailed it: feign "you've never encountered that problem before," provide "the wrong answer or method" initially, then "come upon the best answer."

Yes, just let that sink in for a moment. The most successful strategy would be to pretend to be inept in a way that is believable. We created an interview process in which having the right answer would actually be a disadvantage. That's not a talent filter. That's an acting workshop filter. 🎭

Pattern recognition is the actual job

Here's the part everyone's thinking but no one is actually saying. In most cases, the actual implementation or execution of work is rarely new or unique. That's all there is to it. In addition, the patterns are not infinite in numbers as well. Benchmark data from CodeIntuition.io suggests that just 5 core patterns cover roughly 80% of the interview problems, and CoPilot Interview's data goes further, arguing that the vast majority of problems are "re-skins of about 15 underlying patterns."

Therefore, what you're memorizing is a small deck of cards. And then, you're supposed to act as if you've never encountered it before. The candidate who has is not cheating. They are simply more advanced on the same curve that the job pays.

What real engineering actually tests

If not pattern recall, then what skill is worth measuring? Dan Abramov provided the best response I have come across, during a simulated interview with Theo. His point was that it's not "whether you can reinvent quicksort given a clock."

The cue is: When I give you something unorganized, do you locate the edge where it shatters? Are your decisions sustainable as the subsequent aspect is placed above it? You can't measure that using a timer and a board. The way to measure is to observe someone constructing it. That's probably the reason why OpenAI gives candidates a 48-hour take-home work trial and pays them approximately $1,000 for it, according to an Interview Coder report. It's not a test. It's a simulation of the job. Paying people to do real work and judging the real work. Wild concept.

The honesty trap

So, returning to the initial question: Is it okay to admit that you've come across the problem before? The tension is fake. It only exists because we designed a test where honesty hurts you. If saying "yeah, I know this one" makes you fail, it's not the honest student who's the problem. It's the grading criteria that values performance more than honesty. → Recognizing a solved problem is a feature, not a confession. → A test you can only pass by faking ignorance is measuring the wrong thing. → If your interview punishes the honest expert, it selects for good liars. I prefer to hire someone who can say "I've seen that pattern and here's how that problem fits," over someone who does a slow-motion discovery they memorized the night before. The other one is trying out for a role. 🎬

The takeaway

We continue to pretend that the algorithmic interview tests some kind of 'raw' problem-solving ability. It is in fact mostly testing who has recently practiced those same handful of patterns and who is willing to hide that fact. Reward the right behaviors. Test the skills and the corner cases. And stop forcing the good folks to act as if this is the first time they've ever encountered a coin-change variant. If a candidate solves your problem quickly because they have encountered it before, is that considered cheating, or is that actually the type of engineer you are looking for?