A personal statement from an OpenAI researcher who was never a doomer but has become much more concerned recently. docs.google.com/document/d/e...

Both can be true at the same time: 1. AI progress is still exponential and people who dismissed it were wrong; disruptions will grow 2. 10-15% growth rates and/or extinction-level events* remain highly improbable To be clear, the harms could still be tremendous and we should try to prevent them!

AI will probably be normal technology but it’s also negative EV. AI is inevitable but we can still slow it down. You can hold these beliefs simultaneously.

I don’t really favor AI progress so if sand gets thrown in the gears of the hyperscalers that’s fine, but in practice a lot of the activism against data centers is activism against transmission lines and that has bad implications for wind and solar projects.

I'm very grateful to live in America. Happy 4th, everyone! 🇺🇸🇺🇸🇺🇸

[New blog post] The one celebrity administrator you don’t hear much about is Dr. Oz, the peddler of dubious remedies who now heads the CMS (Centers for Medicare and Medicaid Services). What ever happened to him? chris-said.io/2026/06/03/d...

Dr. Oz as CMS Administrator: A review

A surprisingly competent turn from a peddler of dubious remedies

chris-said.io

New blog post: You don't want luxury unemployment Supporters of AI say that the harms to well-being caused by AI-driven unemployment will be offset by material prosperity. None of the unemployment literature supports this. chris-said.io/2026/02/13/y...

You don't want luxury unemployment

Unemployment causes a profound drop in life satisfaction and material prosperity doesn't matter

chris-said.io

Not to remove blame from Trump, but scientists need to pivot on these association studies. Either the publications come with caveats in bright orange warning labels, or they should stop being funded.

Politico@politico.com · last yr.

Trump plans to announce that using Tylenol during pregnancy could contribute to autism risk and will advise pregnant women to use it only for high fevers.

Nate unfortunately uses some of the same style he criticizes Bluesky for, but I agree with overall point. People here don't realize how alienating this place is to normal voters. www.natesilver.net/p/what-is-bl...

The boring but underappreciated reality is that if you don't want authoritarians taking power, you need to come closer to median voter opinions on crime and immigration.

Matthew Yglesias @mattyglesias.bsky.social · last yr.

Large minorities of the electorate have right-wing views on immigration and crime that mainstream politicians won't embrace, as a result voters back new parties or (in the US) outsider insurgents like Trump. www.slowboring.com/p/a-boring-t...

Large minorities of the electorate have right-wing views on immigration and crime that mainstream politicians won't embrace, as a result voters back new parties or (in the US) outsider insurgents like Trump. www.slowboring.com/p/a-boring-t...

A boring theory of the populist right

A large minority of the public wants tougher policies on crime and immigration

slowboring.com

Seeing reactions to Alpha School on here that feel like the reaction to self-driving cars in 2016. There’s a deep desire for promising results to be fake.

DOGE now doing what it should have been doing instead of firing federal employees randomly for no reason. marginalrevolution.com/marginalrevo...

AI data centers using water is the least of our problems with regard to either water or artificial intelligence. www.slowboring.com/p/theres-ple...

There's plenty of water for data centers

Probably the last worry you should have about either water or AI

slowboring.com

ChatGPT’s water usage is “probably the last worry you should have about either water or AI” www.slowboring.com/p/theres-ple...

[Work opportunity] In the wake of OBBB, Propel is opening up a residency program for exceptional mid-career technologists to help states navigate policy implementation around SNAP, Medicaid, TANF, and WIC.

This might have been the quickest I've gone from "that's an insane conspiracy theory" to "it's plausible".

What makes science credible? Publishing findings like this that undercut whatever political message scientists are supposed to believe in. I am tired of having conversations about "what if my finding undercuts movement x." Tying your findings to politics is what makes science irrelevant.

Elise Cutts@elisecutts.bsky.social · last yr.

New from me for @science.org from #EGU25: The coral reef decline expected due to climate change will enhance the ocean carbon sink — by the end of the century, the seas will absorb extra carbon each year about equal to Australia's annual emissions today.🧪 www.science.org/content/arti...

I witnessed this interaction last week and agree with George. I you want to beat Trump, the way some high profile accounts act here is just incredibly counterproductive.

Many people here say there's no need for Dems to build a bigger tent, since MAGA will destroy the economy and lose elections on fundamentals. Maybe, but I wouldn't bank on it! It's not that hard for Trump to just stop doing dumb things.

If you want to beat Trump, the way you respond to these types of posts is an IQ test.

Many normies are dismissive of AI x-risk, but also think their ChatGPT query will cause water shortages. What should we make of this? (A single California almond uses 300 times as much water).

I promise this is not a defense, but these things are simultaneously true. 1. They’re not playing 11-dimensional chess. 2. The next 5 years are in fact 11-dimensional, so dumb things we do now have an unusually high chance of being good, and vice versa.

“Bhutan is unusually happy” is a popular claim that, in retrospect, had all the hallmarks of being false. x.com/ourworldinda...

Do the terms "credit" and "debit" add anything other than confusion? Why is an increase to an account a "credit" or "debit" depending on which side of the equation it's on? Why not just say "increase" and "decrease?

Going forward, any claims about Elon's mental abilities must specify a year, like Elon[2017], otherwise we will forever talk past each other.

Interesting prediction on AI and unemployment. www.astralcodexten.com/p/links-for-...

One of the creators of the USDS writes about DOGE. "I care deeply about trans people and immigrants [...] but I could not care less that someone plugged in a server to create a new email list without a Privacy Impact Assessment." reason.com/2025/02/13/i...

I'm an ideal target for contrarian takes like "USAID sounds good but is actually bad", but... I just don't understand this? A mix of effective and ineffective international development that is less than 1% of the federal budget seems fine to me.