Welcome back to the AI For Humans Newsletter!

So remember that Columbia student who got suspended for building an AI tool that basically whispered LeetCode answers into your ear during technical interviews? His name is Chungin “Roy” Lee, and not only is he not cancelled, he just raised $5.3 million to take that tool pro.

Meet Cluely, the most Black Mirror-ish tech we’ve seen to date.

Less "Cheating the System," More "Upgrading the Interface"
When Roy’s story first came out, I had some level of sympathy for him. I’m not a professional coder, but I’ve heard how grueling and, honestly, kind of ridiculous those tests can be. Listening to him on Hard Fork, I started to think maybe he was just pointing out how insane our “modern” interviewing system can be.

But this isn’t just about acing interviews. Cluely is pointing toward a future where we might not go into normal, everyday conversations by ourselves anymore. 

The tool acts like a silent co-pilot, feeding you real-time context and suggestions as you navigate work meetings, high-stakes negotiations, maybe even eventually first dates. The teaser trailer points to a future where, likely via augmented displays, we’re able to access real-time suggestions from AI to amplify our conversations, like a futuristic Cyrano De Bergerac, whispering perfect lines in your ear (or Roxanne, if you’re a Steve Martin fan).

The Ethical Line Is Getting Blurrier by the Day
Cluely doesn’t just hint at a future where AI helps us cheat, it hints at one where it quietly enhances us, one whisper at a time. And when devices like Meta’s Orion glasses start rolling out, that whisper could easily live inside our earbuds or contacts.

So then what? If everyone’s walking around with invisible backup, emotional coaching, fact-checking, social cue translation, etc do we just accept that authenticity now includes a co-pilot? Or do we start wondering if every conversation is a one-person play with an AI writer offstage?

Lately I’ve been thinking a lot about AI companions, not the ‘AI girlfriend/boyfriend’ type, but more like an always-on assistant that actually knows you: your tastes, your patterns, your style. And my brain keeps drifting back to the daemons from Philip Pullman’s His Dark Materials.

In those books, daemons are external versions of a person’s soul, spirit animals that walk beside you and guide you, offering advice and insight as you move through the world. It’s starting to feel like AI is going to become something like that. Cluely may seem wild now, but maybe it’s just a rough first draft.

Side note: This season of Black Mirror on Netflix is exquisite and absolutely worth your time. But fair warning, it might make tools like Cluely feel less like clever productivity hacks and more like early chapters in a tech dystopia. The show’s always been great at poking at the tension between human intent and digital power and Cluely, whether it ends up as a helpful sidekick or something more manipulative, feels like it’s walking that same tightrope.

More on Thursday!

-Gavin (and Kevin)

Want to know everything about OpenAI’s new o3 model? Watch the show 👇

We 💛 This - OpenAI’s o3 As Geolocator

OpenAI’s o3 model is here (check last week’s episode for all the details) and one of its most impressive aspects is multi-modal reasoning. This means that it can think about the images you give it and a very fun (and maybe very scary) use case people have figured out so far is geolocation.

It’s a fun little experiment to try today, just upload a photo of your location and say “Can you guess where I am?” and see how close it gets. Or, if you’re one of those people who worry about the AIs knowing exactly where you live, maybe don’t. 😳