Gally

gally.net · Personal blog · on HN since 2019

Gally writes about Japan-based tech experiments, including HN em dash statistics, LLM SVG generation benchmarks, and ChatGPT and DALL·E tests.

What HN saysHN commenters cite his em dash leaderboards to show that humans used em dashes heavily before ChatGPT, and his pelican-alternative SVG benchmark for comparing models.
15HN people shared it
13linked it in comments
6stories submitted
402story points
3 months agolast shared
2019people per year2026

Most-shared posts

  1. 1
    Show HN: Hacker News em dash user leaderboard pre-ChatGPT
    8 people · 377 points · first shared 2025 · 266 comments on HN
  2. 2
    Hn em dash user leaderboard v2
    5 people · first shared 2025
  3. 3
  4. 4
    Standup Comedy by Claude
    2 people · 4 points · first shared 2025 · 1 comments on HN
  5. 5
    DALL·E 2 variations on three common art themes
    1 people · 2 points · first shared 2022 · 1 comments on HN
  6. 6
    Using ChatGPT to Cheat in Class
    1 people · 1 points · first shared 2022 · discussion on HN
  7. 7
    Index
    1 people · first shared 2021
  8. 8
    20230508 syntactic ambiguity.jpg
    1 people · first shared 2023
  9. 9
    20250201sampletranslationprompt
    1 people · first shared 2025
  10. 10
    2012 Which Languages to Teach.pdf
    1 people · first shared 2022
  11. 11
    Sources
    1 people · first shared 2020
  12. 12
    20250304macoswindowopeningposition.jpg
    1 people · first shared 2025
  13. 13
    Poster.png
    1 people · first shared 2026
  14. 14
    Bathing
    1 people · first shared 2023
  15. 15
    20230820hilarystorybyclaude2
    1 people · first shared 2023

What HN people wrote when they linked it

tkgally · 2025-11-25
I added Opus 4.5 to my benchmark of 30 alternatives to your now-classic pelican-bicycle prompt (e.g., “Generate an SVG of a dragonfly balancing a chandelier”). Nine models are now represented: https://gally.net/temp/20251107pelican-alternatives/index.html
acdha · 2026-08-06
You’re misunderstanding the problem: LLMs using emdashes more than the average human still doesn’t tell you that the poster you’re interacting with now is an LLM and not one of the many humans who used emdashes before LLMs existed and are the reason why LLM training picked that up as a good trait. Humans shouldn’t stop writing well just because bots were trained to mimic them. One other problem with this as a heuristic is technical: they used to be hard to enter on Windows so mostly only professionally-edited text there had em-dashes while the ease of entry on Mac, iOS, Android, and to a lesser extent Linux meant that they were more common there. I suspect that this list disproportionately lists Mac users: https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html
mananaysiempre · 2026-05-08
I do. I usually notice and try to rephrase, though. (Also, you can pry my em dashes[1] from my cold, dead hands.) [1] https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard-v2.htm… says mean 1.64, maximum 13 em dashes per pre-ChatGPT comment.
ben_w · 2026-04-27
"Convince the examiner that you are the human" is much harder than "Don't write with the known list of tropes an LLM often uses". I've had my own writing accused of being LLM-generated even here. This may be due to how often I naturally use em-dash, or it may be deeper, IDK: https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html
Someone1234 · 2026-03-10
https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html So if EM-Dash is good proof of AI usage, and people who we can see didn't use AI / or predate AI being popular, are flagged, then that undercuts it by a lot.
latexr · 2026-03-05
> The lists in the post look like they're LLM-formatted, em-dashes etc. No, not “etc”. What else looks LLM-formatted about them? Because em-dashes are not enough to claim LLM. Look, I get that you don’t care about proper typographic characters. You don’t have to, that’s fine. But many of us humans do. https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html And going from “LLM-formatted lists” (without any certainty) to vibe-coded project is a huge leap.
Wowfunhappy · 2025-12-15
Well, (some) people on HN definitely used them before ChatGPT. https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html (And as #9 on the leaderboard, I feel the need to defend myself!)
rl3 · 2025-12-07
>I believe Bryan is a well known em dash addict I was hoping he'd make the leaderboard, but perhaps the addiction took proper hold in more recent years: https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard-v2.htm… https://news.ycombinator.com/user?id=bcantrill No doubt his em dashes are legit, of course.
Freak_NL · 2025-09-30
In Dutch I have the luxury of being able to use the semi-colon in a similar capacity — useful to avoid the em-dash stigma I suppose — but I refuse to relinquish it in English. Fortunately, I can always just point to evidence of prior use here: https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard-v2.htm…
kps · 2025-09-15
Made me look! 265 in _Alice's Adventures in Wonderland_. I use em dashes and semicolons and ellipses and parentheses a lot, although I barely make the leaderboard¹. My doctor says I'm human. In the pre-Unicode era of Usenet and mailing lists I used ‘ -- ’ for a long dash. ¹ https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard-v2.htm…
BHSPitMonkey · 2025-09-15
Then you're throwing out a lot of babies with your bathwater: https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html I'm not prolific enough to rank on this leaderboard, but I often use the em dash in comments/posts/texts and have for years—especially on my phone, since it's easiest to reach from a mobile keyboard.
dang · 2025-08-30
v1 (the submitted URL) was https://www.gally.net/miscellaneous/hn-em-dash-user-leaderboard.html. We've replaced it now v2, for more complex analytical em dash explorations :) - see https://news.ycombinator.com/item?id=45075379 and https://news.ycombinator.com/item?id=45072635.
MacsHeadroom · 2023-04-19
Those are some incredible results! You can do more than ask it to describe the picture. For example: "Write a short fictional story about what the sleeping man in this picture is dreaming about." https://www.gally.net/sleepers/2004-5winter/07.html LLaVA: "In the image, a bearded man wearing a jacket is laying down on a sidewalk, seemingly taking a nap. As he dreams, he finds himself transported to a tropical island where he is surrounded by lush greenery and the sound of waves crashing on the shore. He can feel the warm sunshine on his face and the soft breeze rustling through the palm trees. In this dream, the man is completely relaxed and carefree, enjoying the serene atmosphere and the beauty of the island."

Keyboard shortcuts

/Search
j kNext / previous blog
EnterOpen selected blog
vVisit the blog site
fFollow / unfollow
lSwitch grid / list view
dSwitch dark mode
g hHome
g pTop posts
g tTopics
g fFollowing
g lLiked
g yHistory
g qFAQ
EscBack / close