The Columbia Student Who Sold "Cheat on Everything" for $15 Million

To prove that his software worked, Chungin “Roy” Lee recorded himself passing an Amazon interview and posted the video to YouTube. The software, Interview Coder, was invisible to screen-recording tools and fed him answers in real time. According to CNBC, Amazon and the other companies that had made him offers then rescinded them, and YouTube took the video down after a copyright claim by Amazon. Lee uploaded it again. He told CNBC that he had never planned to work at Amazon, Meta or TikTok anyway; he wanted to show how easy it was to game LeetCode and force companies to find a better test.

He was 21, a computer science student at Columbia University, and by CNBC’s account the university had opened disciplinary proceedings against him. A Columbia spokesperson said the university does not comment on individual students. TechCrunch reported that he and co-founder Neel Shanmugam were suspended over Interview Coder. Wikipedia, summarizing earlier reporting, says the suspension ran one year, and that Business Insider reported it concerned the sharing of disciplinary materials rather than the tool itself. Accounts differ on the reason. CNBC reported that Lee was moving to San Francisco to build the company.

The product was the provocation

Interview Coder sold for $60 a month. Lee said he built it in under a week and told CNBC it was on track to reach $1 million in annual recurring revenue by mid-May 2025, a projection and not a result. His case against the interview was a fair one in outline: “Everyone programs nowadays with the help of AI,” he said. “It doesn’t make sense to have an interview format that assumes you don’t have the use of AI.” Google’s chief executive, in a town hall recording CNBC reviewed, was already musing about returning to in-person interviews.

Cluely, the company that followed in April 2025, widened the idea from interviews to everything. Its launch video, as TechCrunch described it, showed Lee using a hidden AI assistant on a date to mislead a woman about his age and his knowledge of art. In June, Cluely announced a $15 million Series A led by Andreessen Horowitz, about two months after a $5.3 million seed co-led by Abstract Ventures and Susa Ventures. Two investors not involved in the deal told TechCrunch they put the post-money valuation near $120 million; a16z declined to comment, so that number is an estimate. TechCrunch also said the company was profitable, according to Lee’s own posts and podcast appearances.

The number that was not true

In July 2025, Lee said ARR had gone from $3 million to $7 million in a week. In March 2026 he took it back. TechCrunch reported that Lee posted on X that the claim was “the only blatantly dishonest thing i’ve said publicly online,” and called it a “formal retraction.” The figures he posted instead: consumer ARR of $2.7 million and enterprise ARR of $2.5 million. TechCrunch also noted that Lee had described the earlier interview as the result of a “random cold call,” while the company’s PR representative had in fact pitched it to a reporter. Those corrected numbers are Lee’s own and were not independently verified in the piece.

By then the pitch had moved. At TechCrunch Disrupt in October 2025, Lee said “maybe we launched too early,” that Cluely is “not the fastest growing company of all time,” and that the aim was to “become the best AI note taker.” “What I’ve learned,” he added, “is you should never share revenue numbers.” A company that had marketed itself as a way to cheat on everything was now selling meeting notes, in a crowded category, and cluely.com describes an AI meeting assistant today.

What it is evidence of

Three things.

First, attention is now a financing instrument. Cluely’s marketing was built to provoke, and the provocation produced users, press and a top-tier round before the business had settled on what it sold. The investors were paying for distribution as much as for software. That works until the distribution has to convert, which is the point TechCrunch’s November headline made: viral hype is not enough.

Second, a young founder’s own metrics are marketing too. Lee’s retraction is unusual for being public, but the incentive that produced the original claim, in an economy where a screenshot of revenue can move a funding round, is not unusual at all. The reader’s rule is the one this publication applies to everyone: attribute the number to the person who said it.

Third, luck and timing. The interview-cheating pitch landed in a year when companies were actively debating whether remote coding tests still measured anything, and a suspension turned out to be better publicity than an internship. Few students could replicate that, and the fact that the product pivoted away from its founding premise within a year suggests the premise was a launch pad rather than a business.

Lee’s path skipped the Amazon internship. What it produced is a company with a few million dollars of reported recurring revenue, a $15 million round already raised, and a founder who has learned, in public, how expensive a headline number can be.