Just a friendly note that: (1) Sparta didn’t win out in the end; (2) much of what we know about Sparta is probably a lie anyway.

The Trump admin is allowing Anthropic to release Mythos to 100 select companies and also lifting the ban on non-American Anthropic employees from accessing the model. Combining this with the Trump admin also handpicking who gets access to GPT 5.6, government oversight of access to AI models is here

Gather ye tokens while ye may

And people accused those of us on the left of being alarmist when we worried about a Big Brother police state. Troops are now patrolling empty tourist areas looking to arrest people for thoughtcrimes.

The emperor has no reflecting pool

Q: How can we reduce this entire system and its complex, multi-faceted dynamics down to a single number with zero loss in information so that we never have to answer any questions about what's going on under the hood? A: Uh... well that's impossible.

The "standard practice" of only putting footnotes/endnotes at the end of a paragraph makes precisely zero sense to me.

Are AI subscriptions subsidized temporarily or are they going to be persistent loss leaders? Are the per unit prices higher with usage based pricing because SaaS business realities incentivize firms to charge less per unit with subscription price so they can smooth cash flow?

I’ll see that and raise you:

🎶 Swee-eet home Aal-abama!

Got the George Wallace like racism going. Republicans are desperate and doing their very best to rally as many dingbats as they can.

The critique of anthropomorphizing language about AI has gone too far when the mere mention of the term "agent" provokes unearned ire. I've already listed numerous terms of non-LLM-based software from like the 2000s that has anthropomorphic terms.

AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI AI ayayayayayayayayayayayayayayay

YES. Bring back the word “problem.” When I hear “problem” I think “what’s the potential solution if one exists” When I hear “crisis” I think “we are doomed” The former is a better bet and a way to stay sane than the latter.

Job descriptions are endogenous to the firm’s recruiting strategy, so using them to measure the skills that jobs entail seems fraught with danger.

A falling savings rate combined with higher interest rates is not a Good Thing.

I would just cool the entire EARTH down before I cooled a single household down. /s, as I hope is clear

One of the consequences of generative AI is that any open process for submitting content will be inundated with AI slop. This includes government processes asking for comment from the public. There’s startups whose entire business is posting fake AI comments for or against various policy proposals.

Trump’s allegations of a “rigged election” in 2020 are false, but his administration’s attempts at illegal federal interference in 2026 are very real.

Four years ago, the Dobbs decision unleashed a nationwide attack on reproductive freedom and the constitutional right to privacy.

Decisioning and incenting

Honestly, there are federal and usually state laws prohibiting this. 18 U.S. Code § 1701 - Obstruction of mails www.democracydocket.com/news-alerts/...

Trump won't sign the bipartisan housing bill today as planned.

There's no privacy structure in place to manage this highly sensitive data transfer, and the lack of privacy is *by design.* USAID was gutted, and even then routine health/humanitarian data sharing btwn/with states knowingly heightens local vulnerabilities and is a practice of colonial domination.

Listen, I am willing to just accept it and move on, but I don't understand why Ghosh-Lin estimator in semi-competing risks doesn't normalize the incremental count of repeatable events by the *exposure time* instead of total who survived to the beginning of the interval

Feels kind of like everything is about to break tbh

Every new low is just inconceivably bad

Everybody is eating up the narcissism and work from home research and it’s like: Credulousness affects all ideologies.

I don’t think I will ever go to Las Vegas again unless I am forced to.

This is one of those studies where I read it and realized, "I do not need to do good science to become a 'thought leader'."

I got to this study because I saw a NYT headline and I was like ".....how on earth would you study this" Always ask "how on earth would you study this" www.nytimes.com/2026/06/22/o...

huh. thought the strait was open. weird how they're just... abandoning all the ships in the persian gulf by evacuating the crews. wonder what's up with that.

On SpaceX stock value: listen I don’t like the guy either and I am highly skeptical of the company’s valuation, but also it’s been a week and a half so maybe cool it on the victory lap and/or existential crisis.

"An Indiana University instructor who showed a graphic in class that listed the “Make America Great Again” slogan as covert white supremacy has lost her job." www.nytimes.com/2026/06/17/u...

100 percent agreed.

I have not written a peer-reviewed journal article for quite some time, but I do read a whole lot of them and have in some cases applied their results to help companies either make or save a lot of actual money. That said, having not done the thing for a long time, maybe I’m missing something.

I think Christmas Carol and Treasure Island were published as books 40 years apart. Add 40 years to Treasure Island publication date and you’ve got 1923. So I’m pulling for a muppet adaptation of Gibran’s The Prophet. There is no other choice. Clearly.

Have @dingdingpeng.the100.ci and/or @vincentab.bsky.social and company written a paper yet about the models as the prediction machines approach applied to causal inference that looks at the issues of regularization bias, and/or unit-varying effects? (Please do)

In Opinion The system that turned Gila monster venom into GLP-1 drugs is being dismantled through research budget cuts, Jeff Coller writes in a guest essay. “Less support for scientists means strange questions no one will get to chase. Exploring those questions is how medicine advances.”

Who could have predicted this? 🤷🏻‍♂️

Mad king watch

If you are trying to get the first hot takes out in what [insert latest hidden state transition of Straits of Hormuz here], your feet must hurt.

This is all getting a little tiresome

These papers are $&@!ing solid gold. journals.sagepub.com/doi/10.1177/...

Why in god’s name would anyone put carpet on this beautiful wood. LET THERE BE LIGHT!!!!!!!

"The people most worried about AI exposure are in the jobs that everyone has been saying have the highest AI exposure for three years now."

This paper is so $*&(ing good. link.springer.com/article/10.1... Let me explain.

I hereby declare that I wish I were smart/diligent enough to (1) have finished my PhD; (2) applied for and been accepted for some kind of Max Planck postdoc.