February 25, 2026
Dash‑pocalypse on HN
New accounts on HN 10x more likely to use EM-dashes
HN newbies’ dash habit sparks bot panic — commenters go full chaos
TLDR: A small study says new Hacker News accounts use em‑dashes far more and talk AI more, fueling bot suspicions. The community split fast: some say it’s damning, others argue it’s just AI‑assisted humans—and trolls are now spamming dashes to wreck the signal, raising trust and moderation stakes.
Hacker News is in a full-on whodunnit after a user’s mini-study claimed new accounts are nearly 10x more likely to drop em‑dashes — yes, these — and to mention AI (artificial intelligence) and LLMs (large language models). With around 700 comments sampled and data posted, the mood swung from “busted!” to “hold on.” One veteran called it “pretty damning,” while the author added a spicy twist: eyeballing last night showed a wild 32:1 dash ratio, hinting things may be shifting fast.
Then the comments turned into a live experiment-slash-comedy show. One user literally carpet-bombed the thread with em‑dashes to “skew the results,” while another proudly asked how many dashy posts were just “clankers?—like me,” gleefully trolling the metric. Skeptics fired back: can anyone truly tell a bot from a human who used AI to “polish” a sentence? Meanwhile, practicals wondered if these dashy newbies even earn karma (HN’s upvote score) or just vanish. It’s peak internet drama: data, vibes, and memes colliding. The only clear winner? The em‑dash, now the most suspicious punctuation on the site — and the community’s favorite prop for meme‑forensics and mischief.
Key Points
- •The author scraped Hacker News /newcomments and /noobcomments to compare comment patterns of new vs. established accounts.
- •New accounts’ comments used em-dashes/arrows/similar symbols far more often: 17.47% vs. 1.83% (p = 7e-20).
- •New accounts mentioned AI/LLMs more frequently: 18.67% vs. 11.8% (p = 0.0018).
- •Each group’s sample size was about 700 comments, indicating modest samples but large observed differences.
- •Data and code for the analysis are publicly available on GitHub.