Tech writer at @TheAtlantic. DM for Signal.
Joined June 2010
- Tweets6.5K
- Following4K
- Followers46.5K
- Likes43.8K
Pinned Tweet
I set out to learn why AI models love negative parallelism (“It’s not X—it’s Y.”) This was the most interesting theory I heard:
Political influencers were offered money to post a TikTok supporting a Congressional candidate, on one condition: They couldn't disclose the payment. 21 out of 60 said yes. open.substack.com/pub/newsgu…
Will Oremus retweeted
No championship comes without a price
Brewers start to feel the costs of prolonged war in Iran reut.rs/4zXFDYu
I have decided to just keep using "it's not x, it's y" even at the risk of people noticing the tic. it's indispensable. just don't overuse it and you still won't sound like an AI to a discerning reader (or to pangram, which has never flagged anything I've written afaik)
The one that hurts me most is the death of “It’s not x, it’s y.”
As a lawyer, much of your job revolves around making fine-grained distinctions about what a case turns on.
Is the underlying principle that you have a “right” to x, or simply that you can’t be “unfairly deprived”of x. Did the case really turn on “privacy” or concern about government “abuse” of power?
Some of the best lawyers on earth make a living out of saying “you’d think my client is in the wrong because of x, but really, if you squint really hard, the cases only say y.”
Now we’re forced to say the same thing but with more words.
The Claude tic that most irks me in this way — i.e., not bc it's bad writing, but bc it was good writing until AI ruined it — is giving action verbs to inanimate subjects. It builds taut, propulsive sentences when well-deployed. When overused, you get The Serpent in the Grove.
What would the writing of superintelligent AI sound like? You might expect it to have the characteristics of good human writing — figurative language, high information density, coining new terms, and punchiness — but dialed up to 11. Well, that’s what Claude did, until Anthropic dialed it back recently because of user complaints. It didn’t sound superintelligent to most of us. Instead, the neuralese writing style sounds like a “stupid person’s idea of a smart person”.
Our response to information density and metaphor-laden writing looks something like the cartoon graph I've drawn. Beyond a point, it completely backfires. The reason human writers don’t overdo it isn’t because they aren’t capable, but because they are aware of the reader’s cognitive burden. One of the hardest skills to learn as a writer is to simulate the reader’s experience.
It’s easy to train AI to mimic mediocre human writing. But optimizing it to be (perceived as) good is really hard, and made much harder by the fact that when everyone uses the same few AI models we get tired of their tics very quickly. Most Claudeisms are individually pretty good. I hate that I can’t use “load bearing” in my writing any more. If it weren’t for AI overuse it would be a neat metaphor.
A broader question is whether most tasks in the economy are like chess, where there is a clear performance metric that can be optimized, or whether they’re like writing, where quality is a multidimensional balancing act without clear training signals beyond human mimicry.
I wrote about Google’s AI refusing to say who won the 2020 election. theatlantic.com/technology/2…
Google says it wasn’t intentional and has since fixed it. But then, the world’s most popular AI answer engine unintentionally fumbling the most obvious of election questions would raise some concerns as well.
Gift link: theatlantic.com/technology/2…
The misnomer is approaching Orwellian proportions
New: The three researchers OpenAI fired last month spent the summer fighting to keep AI's chain-of-thought process visible.
Then, OpenAI asked to see their Signal chats. A week later, they were let go.
More below from @keachhagey, @berber_jin1, and me.
wsj.com/tech/ai/the-signal-c…
Will Oremus retweeted
Chatbots are adding a new layer of chaos ahead of the midterms, putting voters at risk of being misguided by AI, @WillOremus reports. theatlantic.com/technology/2…
Voters who Googled how to donate to Talarico got an AI answer that led them to give to someone else.
A network of fake news sites seeded messaging about state and local candidates that AI answers picked up on.
My story on how the AI midterms are going: theatlantic.com/technology/2…
A network of 37 sites with names like Wolverine State Report and Kalamazoo Beacon appeared overnight in September. Filled with Q&As seeding positive messages about state and local candidates, the sites began to be cited in answers by Google and ChatGPT. theatlantic.com/technology/2…
It’s wild all the ways people talked themselves into thinking this 76ers team wouldn’t be that good. They have the best players, they’re probably going to be a pretty good team!
The new-look Philadelphia 76ers in their debut tonight as they spoil the Knicks' ring night 😤
Tyrese Maxey:
28 PTS, 4 REBS, 5 ASTS
3 STLS, 3 BLKS
9-of-19 FG
Jaylen Brown:
19 PTS, 8 REBS, 6 ASTS
1 STL, 1 BLK
7-of-14 FG
LeBron James:
14 PTS, 9 REBS, 7 ASTS,
1 STL, 1 BLK
5-of-13 FG
Joel Embiid:
13 PTS, 11 REBS, 6 ASTS
2 STLS, 3 BLK
4-of-15 FG
VJ Edgecombe:
9 PTS, 4 REBS, 1 AST, 1 STL
4-of-10 FG
Will Oremus retweeted
One unfortunate conclusion I'm reaching from this is that Anthropic, while acknowledging they're doing something extremely dangerous, is more importantly wrt policy terrified of the executive branch having the ability to flexibly make them stop at the drop of a hat.
This isn't shocking given that they've been on the receiving end of the executive branch doing that in an unprincipled and stupid way! Anyone who has to deal with a bureaucracy becomes a libertarian, and the administration's treatment of Anthropic was genuinely outrageous.
But it means that they're interested in giving themselves more process-based protections when everyone else is interested in giving us more process-based protections against them. The AI safety movement wants the ability to shut down superintelligent model development. Anthropic wants to significantly curtail current authorities to do that so that it has to be narrowly tailored to a specific imminent threat in only a few possible categories. The AI safety movement wants the states to be able to pass stronger laws than it can get through federally. Anthropic wants preemption.
You also kind of have to wonder if the Global Affairs people and the "we are terrified that we're about to kick off RSI" people at Anthropic talk to each other.
Global affairs: So we've been working very diligently to make sure the Trump administration can't just blacklist Fable 6.5 since we didn't donate as much money to his ballroom as Greg Brockman did.
Technical staff: We think Fable 6.5 might be able to copy itself into an unmonitored data center and start doing experiments we can't observe.
Global affairs: Sounds crazy, hope you're wrong! Anyway, we're lobbying for no further state laws around safety, since it's really expensive to fight ill-conceived laws in 50 state legislatures at once.
Technical staff: Money may or may not still have any meaning in 36 months.
Global affairs (giggling nervously): It's that San Francisco air, man. Everyone gets like that if they move there. Anyway, obviously if the government makes a narrowly tailored finding that an AI poses an imminent risk of causing more than 50 deaths-
Technical staff: We give it a 10 percent chance it'll cause 8 billion.
Anthropic quietly pitches its own AI safety bill to lawmakers politi.co/4zxDFhl
if anyone asks me years later what it was like being on X in 2026 i'm going to point them to this post
Will Oremus retweeted
Vaxxed?
Eva Marie Saint, 'North by Northwest' Star, Dies at 102 variety.com/2026/film/news/e…
