VIEWS5.9K
Derek Thompson@DKThomp
Not a dig at @TheZvi, just a good news peg for the observation
2hViews 5.9KLikes 7
Original post unavailable.
Sentiment
Many users criticized assuming AI authorship for longform tweet essays because it produces slop that disrespects readers and undermines valuable signals, while a few recommended experts on the topic.
Pos
11.1%
Neg
88.9%
9 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
BOOKMARKS1
Daryl Fairweather, PhD | Chief Economist@FairweatherPhD
@DKThomp Selection bias. People only post the AI tests that come up positive
2hViews 131Likes 4Bookmarks 1
LIKES14REPLIES2
Alan Cole@AlanMCole
@DKThomp I don't regret the creation of AI in the least overall, but an old heuristic, the "effortpost," where length suggests somebody actually really cares about an issue and spent lots of time putting together their points, is no longer valid.
2hViews 448Likes 14
RETWEETS1
Password is Taco@pw_is_taco1
@trumwill @AlanMCole @DKThomp Yeah, every year you see at least one story like this that is supposed to be heart-warming, but all I see is a massive blackpill about the admissions process
2hViews 59
Will Truman@trumwill
@AlanMCole @DKThomp Sort of like applications, where before anybody could send them out by the 100's, the gatekeepers (HR or college admissions) would look at them. That system doesn't work anymore.
2hViews 38Likes 3
Zvi Mowshowitz@TheZvi
@DKThomp How far would you extend this into other forms? This was crossposted from an official Department of State substack, where a lot of the posts are by Rubio.
But yeah, if it's tech bro longform article on Twitter it's going to be some form of slop most of the time.
2hViews 85
Candyman@sugarbets_
It’s funny because in sales and marketing it’s well understood that more information *can* automatically make your product seem more valuable.
“Scroll ‘til sold” pages come from that.
Not super common anymore, but there’s also a psychological factor with longer writing seeming more important or more valuable
2hViews 19Likes 1
Alan Cole@AlanMCole
I'm using "frontier" liberally here, by the way.
Lots of important jobs work on some sort of knowledge frontier. I don't mean like cancer research necessarily, but at least something like the knowledge frontier of "recognizing credit card fraud today for some sub-set of transactions US banks deal with."
There are obviously some principles and heuristics that are known already and uncontroversial and unremarkable and available on the open internet. But the field is also ever-changing with conditions, and it's genuinely valuable to the point that businesses pay people to work on the cutting edge of it, not just to know the basics.
Pre-AI, if you had a guy who personally to took the time to write a cover letter to JP Morgan Chase explaining that he reads the equivalent of Patrick McKenzie's blog, and knows issues A B and C and thinks they're interesting, he'd probably get a job interview, and rightfully so. That's a big signal.
AI can replicate all that in a few minutes. So now people have to find other signals.
2hViews 16
Alan Cole@AlanMCole
To generalize:
It used to be the case that "sub-frontier but pretty good knowledge, plus executive effort and polished presentation," was a GREAT signal of someone's potential.
Unfortunately for the signaling mechanisms we used in the past, AI excels at "sub-frontier but pretty good knowledge, plus executive effort and polished presentation."
2hViews 14
Derek Thompson@DKThomp
@TheZvi ya good point Substack seems to have a lot of AI writing too
2hViews 63Likes 1
G, MD@DrBeavisAI
@DKThomp So people use more AI but then AI fans complain they use AI to write? Seems illogical.. Soon it will be socially acceptable to have all long form texts written with AI (with good input)
2hViews 148
Vincent Murphy@brainstormvince
@DKThomp apropos not much but interesting 'long form' as a term only coined post internet/social media almost a deliberate attempt to differentiate 'printocene era' considered / depth writing from fluffy digital stuff
2hViews 89
Andrew Harvey@AndrewRHarvey
@DKThomp This is why I write short multi-message tweets. With occasional profanity.
2hViews 76
DoTheMath@_Do_The_Math
@DKThomp And then no one actually reads them -- at best they use the same AI to summarize it into a couple bullet points.
Wonderful world we've created ...
2hViews 74
Confection@ConfectionTV
@DKThomp Unfortunately, the same assumption is also more and more accurate with student papers
2hViews 50
Ben Horne@benjamin_horne
It’s telling that X Articles are pure AI slop literally 90% of the time.
While there’s lots of amazing people/accounts that keep me coming back here (and X admittedly has amazing distribution powers for writers, especially new ones) the reality is this app has become super grifter-coded, as a general rule.
You can just grift and slop your ass off on here and you wind up with more supporters than haters.
2hViews 39
employee@browsingatwork
@DKThomp @gwern is the best person to read on this subject
2hViews 34
Xavier Clegg@xavier_clegg
@DKThomp I didn't use any AI to write my newest longform tweet essay 🔽
2hViews 7Likes 1
Kai Samuelsen 🧦🗽🚅@KaiDaigoji
@DKThomp Why do you accept the premise that AI detectors work?
2hViews 16
Michael Quintos@mquintos
It’s funny — and true — that in sales and marketing, more information can automatically make your product seem more valuable. Not because customers always read every word (they often don’t), but because the presence of detailed, thoughtful content signals quality, expertise, and confidence.
This principle powers one of the classic direct-response tactics: the “Scroll ’Til Sold” sales page. These are those epic, long-form landing pages that start with a bold headline, then unfold layer after layer of benefits, features, testimonials, case studies, FAQs, guarantees, and more.
The idea isn’t to bore people — it’s to give them everything they need to overcome objections and feel completely informed before buying.
2hViews 14
Load more posts