People using an #LLM to write a whole novel for them are like the people using #chess engines (like #Stockfish) to cheat in online games. Because both do all the actual work for them, what’s the fucking point?! I don’t mind people using either of them to learn—because that’s what I’ve been doing this past month—but using them to replace the joy of writing or discover-attacking your opponent into a queen promotion on move seven.

I’m not kidding; this was a real game. Bots fall for the strangest traps:

[Date "2025.07.13"]
[White "Jikaxav"]
[Black "Sven-BOT"]
[Result "1-0"]

1. e4 c6 2. d4 d5 3. exd5 b5 4. dxc6 e5 5. Bxb5 e4 6. c7+ Qd7 7. cxb8=Q Rxb8 8. Bxd7+ Bxd7 9. f3 Rb6 10. fxe4 Rd6 11. Nf3 f5 12. e5 Rb6 13. O-O Nh6 14. Nc3 Rc6 15. d5 Rb6 16. Qd4 Ke7 17. Qc5+ Kd8 18. Bg5+ Ke8 19. Qc7 Rxb2 20. e6 Bxe6 21. dxe6 Bc5+ 22. Kh1 Kf8 23. Qd8# {1-0}

#chess #pgn

And yes, I do analyse all my games after I’ve played them. Because Chess.com wants you to pay to win, I just export these to #Lichess or to Stockfish on my computer to get a “technical report”, and then to a really, really long #ChatGPT conversation (two of them now, in fact), which has essentially served as an impromptu chess coach. Chess is 99% tactical pattern recognition and so are LLMs, so perhaps I’m onto something here.