Cluely, an AI app that promises to help you “cheat at everything,” perhaps deserves some credit for honesty.
If you believe that artificial intelligence is a platform that, in some ways, can help you game the system by delivering melifilous prose written by a large language model (LLM), stunning art generated by ChatGPT, or convincing video percolated by Sora, then maybe Cluely and its founders are just being honest.
It’s more likely, though, that they are blustering their way to the launch of a product that makes a mockery of integrity.
Cluely is a crafty app that installs in your browser and then can sneakily watch and listen to everything on screen. The primary use case is a Zoom video call where you’re being interviewed for a high-paying development job. During such a call, you might be asked to Leetcode, which is essentially solving coding or algorithmic problems in full view of your interviewer. With Cluely running, it could quickly tell you how to respond to questions and even solve the coding problems.
Lee claimed to have used Cluely to breeze through interviews with multiple major tech companies. Word apparently got around to Columbia University, where Lee and his Cluely co-founder were students. Lee was, according to letters he posted on X (formerly Twitter), kicked out. Lee detailed the entire experience in an X thread.
Instead of remorse, Lee seems to express a level of satisfaction, like this is a plan gone well. And, in a way, it is. Cluely has, according to Techcrunch, raised $3.5 million in funding and is prepared to help us “cheat at everything.”
If that statement makes you feel a bit nauseated, you may not be alone. Lee has spent a fair amount of time defending the app on X, and Cluely may not have helped its case with a tongue-in-cheek launch video that depicts the 21-year-old Lee using Cluely in a smartglasses-like app to convince his date that he is 31 and interested in her and her art. She catches on to the subterfuge and ultimately walks out, but the point is made. Cluely could help you cheat at dating.
While Cluely has yet to reveal which LLMs it uses to power its answers, this is a funhouse mirror view of AI’s potential.
In his manifesto (which, it appears, has since disappeared from the site), Lee argues that we need to redefine cheating and that what Cluely’s doing is no different than Google or a calculator, which put answers at our fingertips and perhaps have been helping us “cheat” for decades.
On X, I asked Lee about this perspective, writing:
“Do you worry that your experience teaches the wrong lesson about AI? People are already skeptical, and if they see it as a tool for cheating, maybe some will be less likely to use and engage. At the very least, more people will trust less (on things like job interviews) because they might believe the applicant is using AI to cheat.”
Moments later, Lee responded (For readability, I’ve fixed the punctuation):
“What exactly does “cheat on everything” mean?
How do you ‘cheat’ on a conversation?
Cluely (and AI) enables such an insane amount of leverage that it feels unfair, and “like” cheating.
The world will be uncomfortable with this reality (as you yourself show), but it will soon become normal, and the day that AI maximalism is truly normalized, humanity’s potential will 1,000,000x.
Initial hesitation -> massive adoption has been how every single technology in the history of the world has spread.”
Cheating is…cheating
While I can’t disagree with the notion about hesitation and then mass adoption, and am seeing that in real-time with consumer-grade AI, Lee seems to imply that you can’t “cheat” at a conversation. I guess he has never read Cyrano de Bergerac.
None of this explains away the fundamental deception coded into Cluely.
I followed up on X with:
” ‘Cheat’ is a loaded word, but I also think it’s one we do understand. Google search is super-powered research, a digitization of encyclopedic knowledge at your fingertips. A calculator is a math wiz in your pocket, but one that was banned from class unless educators allowed it. They understood the potential for cheating. A browser tool that watches and listens and then produces an answer without the knowledge of the convo’s other participant is, many will agree, something different. BTW: Which LLM are you using?”
Lee has yet to respond, but perhaps that’s because he knows the answer. Google and Calculators are tools. Cluely is a tool hidden behind a lie: “This is me answering.”
It would be the same as secretly using Google or a calculator to answer an interviewer’s questions. There’s a fair chance they would notice your fingers frantically typing just off camera. Cluely just supercharges the subterfuge, but it doesn’t make it any less like cheating.
A brief hands on
I did, by the way, install Cluely on my MacBook Air to see how it works. It’s a powerful tool that overrides, with your permission, normal safeguards that stop third-party tools from watching the video and listening to the audio on your screen. It offers no alert to the person on the other side that they’re being watched by a massive AI brain.
It’s early days, and Cluely does not appear to be fully real-time. You have to click the AI button for it to work and provide answers. Even so, the potential is there, and it’s not necessarily a good one.
Lee’s efforts to normalize and redefine cheating won’t make it any less of a deception or any more acceptable. He got kicked out of school for it, and now he wants you to put your own reputation on the line. Some might do it, I know I won’t.
You might also like
