It's kind of funny watching environmentalists who definitely aren't influenced by Tom Steyer's climate money, fawn over him for governor (I'd probably vote for him if I lived in CA)

I know you're on too and I will say I can totally see the mechanism for them being effective just by being a way of controlling food noise.

Will note that bulimia was actually a condition heavily improved by GLP-1s according to a meta-analysis. Even on its own terms I'm not sure it works. www.nature.com/articles/s41...

This is an all timer from the cut.

Relative productivity of construction is way down.

I'm actually pretty curious what Anthropic's numbers look like. Obviously OpenAI is obviously just spending absurd amounts but Anthropic has raised $69B to date, and is clearly quite conscious of its subsidy levels right now. S-1s will be interesting.

Right. The number of times I've said "I feel like the law should be X but I've never seen a case that really says it" and then tried to write it and hoped I could find a case to support it. Now I can just write the sentences and say "is this proposition correct?" and get helpful answers.

Feels like not a big deal. For API clients I assume the marginal cost of a token for Anthropic is priced below cost which wasn't even the case for Uber right? It's just all capital costs and eventually the music can stop.

Personally I think being a litigator is going to be a lot more fun going forward. Doc review, annoying research, drafting discovery requests, all so much easier. The fun stuff is the hardest to automate.

More and more people are saying. So much of the job of research in the law is "i can describe the thing i need to find pretty well but it's actually pretty hard to find the specific words that might get used and how" and LLM vector space is basically perfect for this sort of task.

I realized about Ossoff today that one of his strengths is that he does Obama schtick but not an impression. Think there’s an older millennial reverence that’s different than say Shapiro trying to steal his flow.

Donarumma getting in on the action

Gibbs-White and Eze taunting Spurs all season after the transfer saga leading to Spurs getting relegated. Just obvious Spurs made a deal with the devil to win one trophy last season.

The Rockets took 27 more shots than the Lakers last night (including free throw adjustments). How often do you do that and lose?

Businesses can’t afford to stay with all these new customers around. Do you hear yourself?

Making it easier to use psychedelics to treat PTSD is good though.

From, uhh, personal experience, if a Blackhawk is randomly over your house, worth considering if the President or Vice President is in town for a fundraiser.

There are also certain tasks where you don't really care about the quality of the writing. Do I care that my discovery demands have nice sounding prose, or do I care that they're comprehensive and ask for everything I need?

It's also why Claude Code is really good at helping people who understand a game like soccer create a model (a la @themlg.bsky.social) but it is not good at the data science itself particularly. AI has coding chops but not model feel.

With enough scale and the right structured data, AI will be able to provide great outputs for deterministic problems. Coding is great for that because it either works or it doesn't at the end. Whether a sentence sounds good not so much and well a lot of the writing sounds like shit.

The most important way to understand whether AI will be good or bad at something is whether the output is deterministic or not. ML was first used on games because the output had a score, perfectly deterministic. Attention, i.e., how long somebody watches Netflix, is deterministic.

Because of age x feelings towards Israel crosstabs among Jews special elections will overstate the effects but Democrats are just going to lose votes among Jews in 2026 I think particularly in comparison to national swing.

Think it's pretty straightforward!

can spoil it for you: yes

This is something @add-hawk.bsky.social talked about with philosophy too where papers got more and more technical because people got ranked and measured and journal prestige ranked you. I think it’s probably true there’s a lot less bad academia out there now but some real loss too.

You made political scientists learn math and make falsifiable predictions and as a result no one can put their dick on the table and say “here’s how the entirety of markets and society work” anymore without putting it behind layers of abstruse theory abstraction

The incumbency advantage has also gone down over time. Broadly the news environment is just more negative. That may have, probably has, gone up in the past few years with even tighter algorithmic delivery but still people just really fucking hate inflation and 2022 was a massive shock.

While I think the Michigan survey needs adjusting and I think the vibes hypothesis is quite overrated, broadly news and sentiment. POTUS approval systemically lower, news has gotten more negative driven by partisan news sources.

Yes this is totally right and I think reflected in election swings by young people being much stronger than by older people. If you were 18 when Reagan was re-elected, you were 58 and had not seen annual CPI above 4% once since 1992 and then it was 8% in 2022. Just a massive shock.

Somebody should try to figure out the social benefit of buying out the patent