937

LESSWRONG
LW

936
NewslettersWorld Modeling
Personal Blog

7

Halfhaven Digest #4

by Taylor G. Lunt
14th Nov 2025
3 min read
0

7

NewslettersWorld Modeling
Personal Blog

7

New Comment
Moderation Log
More from Taylor G. Lunt
View more
Curated and popular this week
0Comments

My posts since the last digest

  • Asking Paul Fussel for Writing Advice — I gave AI the works of Paul Fussel, Christopher Hitchens, and Eliezer Yudkowsky, and asked it for writing advice. I got some actually good advice and was surprised the experiment wasn’t a failure. I have since used this trick again to get feedback about my subsequent posts.
  • Halloween Tombstone Simulacra — Noticing the drift between Halloween tombstones and actual tombstones.
  • Minimizing Loss ≠ Maximizing Intelligence — A higher-effort post describing why I think LLMs and self-supervised learning as a whole are dead ends and won’t get us to superintelligence. And some approaches I think are more promising.
  • Turning Grey — A sci-fi story in 2025 that isn’t about AI?
  • I Read Red Heart and I Heart It — A review of Max Harms’ latest novel Red Heart. My post was liked by Max Harms, which I think pretty much makes me a published author now.

I’ve been busy lately. I’ll admit, the Halloween post was a vapid idea I thought of just to get something out quickly. But I think it turned out alright anyway. I am most proud of the short story this time, which I read aloud to my girlfriend and she liked. The Shirley character in the story is literally just my girlfriend, by the way, down to her profession and the way she dresses at work.

Some highlights from other Halfhaven writers (since the last digest)

  • roundness of numbers is complicated (April) — Contra Inkhaven resident Signore Galilei, April of Apriiori cleanly argues essentially that you can’t describe what we mean by the “roundness” of a number with a formula, because e.g. 25 is rounder than 30 when dealing with cents, but not when dealing with seconds.
  • We write numbers backward (lsusr) — I started this fun video thinking, “no we don’t”, and ended it thinking “we totally write numbers backward!”
  • E-Prime (Lorxus) — An overview of an interesting, restricted form of English with the worst, cringiest name ever. I definitely be a “to be” user and have no plans of paring back my usage of the King’s verb, but I appreciate the thoughts about what kinds of language can be unclear.
  • Husky Syndrome (Aaron) — On the mindset of social anxiety with a brilliant analogy to sled-pulling dogs.
  • Supervillain Monologues are Unrealistic (Algon) — Real-life villains monologue endlessly about what they plan to do, and nobody listens. Startup founders, on the other hand, are anxious to tell people their master plans, for fear someone will think they’re foolish (or steal their idea). But nobody will listen anyway, so feel free to monologue as much as you’d like.
  • [The Mortifying Ordeal of Knowing Thyself] (Philipreal) — Contrary to the grandiose title, it's a relatable blog post about nervousness when posting Halfhaven blog posts, and a desire to do less than your best so nobody can judge your true best. It suggests a strategy of posting the occasional higher-effort post, which I have been following myself (two out of my last five were higher-effort).
  • I Admit, I Am Ignorant of Many Things (keltan) — An ode to saying “I don’t know”.

Since the last digest, Inkhaven proper has started, and we off-brand Halfhaven writers are now in competition with the 41 Inkhaven residents for LessWrong upvotes. I’m not including Inkhaven posts in my digests (God knows Inkhaven residents have enough support — they even have a ball pit!), but I’ve been reading some of those as well, and you should check them out. At the beginning of November we also had a few more people join Halfhaven, bolstering our numbers against the Inkhaven hordes. We also had our first early-finish, with Algon writing their 30th post on November 2nd, and now going for a high score (they’re currently at 36 posts).