News for hackers. Stuff that matters.
Re:I had AI read my field by jdlshore on Wednesday November 18, 2026 @01:08AM (#50542443)
The landing page text seems pretty obviously AI-generated, in that prideful way AI regurgitates its instructions. Didn't feel like it was worth reading further.

"158,059 papers and 554,670 citations. Every paper with full text screened, 659 of the most important recomputed and checked, and every verified error followed through the papers that cite it. This is the map, and what it says about whether science corrects itself."

Re:I had AI read my field by sudo_cowsay on Wednesday November 18, 2026 @01:08AM (#50542444)
I am very sorry. My computer cannot handle the website (lag). However from the first few scrolls that I looked through, I don't seem to see any mention of what AI model the author used. Is there any mention of that? Sorry!
Re:I had AI read my field by swampdeer on Wednesday November 18, 2026 @01:10AM (#50542457)
web page that runs at 3fps in firefox
Re:I had AI read my field by joe_the_user on Wednesday November 18, 2026 @01:12AM (#50542471)
I read roughly the whole page. Nice picture and no information that seemed at all useful. It means "the 4% of the papers that mattered". What does this mean and what does it, in turn, matter?

It definitely oozed the feeling of AI pseudo-knowledge.

Re:I had AI read my field by lexandstuff on Wednesday November 18, 2026 @01:15AM (#50542494)
It's funny how Claude-isms change across models, but each new set of quirks becomes so quickly apparent in everything it outputs.

Opus 5.5 loves to talk about things that "nobody" did.

- "Errors that nobody fixed"

- "Nobody told the layout what numerical relativity is."

- "Nobody re-derives every equation they cite"

Also, just an uncanny knack for adding an exhausting amount of detail into the prose:

"every paper I could get the full text of got screened: 50,269 papers, 421 million words, checked for the fingerprints of the worst kinds of trouble. That covers 76% of the papers drawn here. Nothing before 1991 is on arXiv, and journal-only papers aren't open"

Re:I had AI read my field by spongebobstoes on Wednesday November 18, 2026 @01:15AM (#50542497)
this looks like great work! interesting visualization and writeup

I wish it was done for every field. seems like there is a lot of low hanging fruit just in terms of identifying and correcting errors. and then there's the extrapolation work hinted at

it does feel like a great time to be in academics! a few grad students with $200 AI subs can probably have an outsized impact