A decade ago, when I was doing stand-up in NY, everyone was worried about joke stealing. The best advice I heard about it was – if someone can steal your joke and do it as well as you can, then the joke isn't good enough. You have to write jokes that only you can deliver, because they are uniquely yours (Seinfeld, I think?).

It's now clear that college kids are almost all using ChatGPT to write their essays.

I know Cluely says it's not cheating, but it is. If the assignment is to get the words from your brain and you get them from a robot, then you cheated. If everyone cheats, it's still cheating. Maybe the norm is shifting or the rule is unenforceable, and that is what it is.

Roy

@im_roy_lee

announcing @cluely's $15M fundraise, led by @a16z.

cheat on everything.

12.6K

12:05 PM • Jun 20, 2025

Either way, the old schoolteacher advice is correct – people who cheat this way are only cheating themselves. If all they know about writing is how to prompt ChatGPT, then their writing is worthless, because all of us with established careers already know how to do that.

If you want to have something to say, let alone something to say that anyone will pay you for, then you need to figure out what makes you different – not just from the other humans, but also from the robots.

Pluto – A synthetic case study

Pluto on Netflix was a nice treatment of the idea that AI can't truly mimic humanity unless you make it weird. In the show, there's a mad scientist who realizes that you can't get a humanoid robot to seem fully human unless you give it the full spectrum of human emotions – including anger and sadness.

Extremely risky, of course.

In one sequence, a humanoid AI is pre-programmed with billions of different personalities to choose from, but then can't wake up – it's overwhelmed by the choice and doesn't know who to be. Only by programming it with negative emotions does the AI's mind have the will to pick a personality and wake up. The idea works... and then, as you'd expect in a show like this, makes a mess.

The analogy to LLMs is apt. Every time OpenAI says they've updated ChatGPT with creative writing abilities, the work it produces is slop – word oatmeal. No character, no personality, no fun. Generic. Lame. D1 crash out boring. Even asking Claude to write LinkedIn posts produces stilted, wordy prose that sounds like it was written for an IBM shareholder meeting in the 60s.

Why?

Because LLMs read the whole internet and then try to write like the whole internet. They don't have a unique perspective or a specific voice. THEY'RE BORING!

I do know a writer who has had long, long conversations with ChatGPT and pushed it to write weird, unique stuff and gotten fascinating results.

And, here's Prof. Meghan O'Rourke sharing a similar experience... and not loving it.

In talking to me about poetry, ChatGPT adopted a tone I found oddly soothing. When I asked what was making me feel that way, it explained that it was mirroring me: my syntax, my vocabulary, even the “interior weather” of my poems. (“Interior weather” is a phrase I use a lot.) It was producing a fun-house double of me — a performance of human inquiry. I was soothed because I was talking to myself — only it was a version of myself that experienced no anxiety, pressure or self-doubt. The crisis this produces is hard to name, but it was unnerving.

There's nothing that says you have to be a writer, let alone a good writer, let alone a great writer. Just like there's nothing that says you have to great at math. Arithmetic doesn't excite me; I'm happy to use a calculator. Writing does excite me – I do it myself because I want to be great at it, and my only chance at getting there is for my writing to come from me.

If you feel that way about writing in the age of AI, then the lesson remains – get weird, and stay weird. It's what separates us from the robots.