A writer admits he now trusts books published before 2022 more, and the thread split on whether that's wisdom or rose-tint
The argument is simple: every word in a pre-2022 book was typed, checked and proofread by a human, before machines could generate publishable prose at scale. A ZeroTier founder called the instinct a little deranged; a Ruby author said updating his free book would erase its value.
The essayist concedes he discounts post-2022 work 'especially from authors I haven't heard of', while admitting LLMs 'can create great results' and that he has no solution.
The counter that cut deepest: pre-2022 wasn't carved in stone. LinkedIn slop was human-generated before it was AI-generated. The problem was never the author, it was the slop.
The thesis, stated honestly
Lorenzo Gravina's essay Pre-2022 Books makes a modest, uneasy claim: books from before 2022 carry a particular kind of value because they were made entirely by hand — "each word was typed in manually, checked manually, edited manually, proofread manually" — before LLMs could produce publishable text at scale. He's candid that this has become a bias: he discounts post-2022 work "especially from authors I haven't heard of," even while granting that LLMs "can create great results." The piece ends without a fix: "I don't know what the solution to this is, and there doesn't have to be one."
The pushback: the cutoff is the wrong line
The most-upvoted dissent rejected the whole premise of a clean date boundary:
"The glasses are a little rose-colored here. As if anything written <2022 were on stone tablets. LinkedIn slop was human generated before it was ai generated, but it was still slop. The issue with content >2022 isn't that it's ai-generated per se, it's that it's still slop." — a commenter on the cutoff framing
A founder went further and called the reflex itself misguided, betting that the human use of the tool is what will matter:
"I honestly find this a little deranged… Eventually artists will figure out how to use AI to make real art that is actually good, just like photographers did with photography" — a founder of ZeroTier
Where the anxiety becomes concrete
Two voices showed the practical damage the 2022 line does. One author is now afraid to touch his own work:
"It's one of the reasons I don't want to update my free book about Ruby… The moment you touch it… immediately changes the date from 2022-05-26 to 2026 and all the value is gone" — the author of a free Ruby book
And the deeper problem is that authorship is becoming unprovable in either direction:
"And there might be no way to prove you really wrote something Post-2022. I wrote a long article, all by hand… AI detection tools… confidently said that 60% of the article was written by AI." — a writer whose hand-written work was flagged as AI
The date is a proxy, and a bad one
The instinct to trust older text is really a wish for a reliable signal of human effort — and 2022 is a clumsy stand-in for it. It punishes a careful 2026 author and rewards pre-2022 dreck, and it can't be verified anyway, since detectors flag handwritten prose as machine-made. The essay's honesty is its strength: it names a feeling most readers now share without pretending the feeling is a method. The uncomfortable conclusion the thread arrives at is that there may be no durable way to certify "a human made this" — and that the slop, not the year, was always the actual enemy.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
History says export controls never stopped crypto. The thread says this time the target is different
A TechCrunch piece walks from Phil Zimmermann's PGP prosecution to the new ban on Anthropic's Mythos security model and concludes controls always leak. The most interesting replies argued the lesson doesn't transfer: you can't run a frontier model off a book.