Toronto Tower Triumph, GLP-1 for Everyone?
Multi-agent AI framed as a token hack, OpenAI’s math boast meets Tao’s skepticism, and Caltrain ridership climbs
I saw Superman (2025) this evening. It was pretty good.
I have not yet read this article, but I'm happy he wrote it, the title of the article asks a question I want to know the answer to: "If GLP-1 Drugs Are Good For Everything, Should We All Be on Them?"
This is a new meme format, not necessarily the best use of it I've ever seen, but I think the speech bubble wrapping around the retweeted post is a great trick. Obviously you don’t always have to make it the California Almond Board.
I'm surprised that the project is going to be predominantly office considering the high office vacancy rate in SF, but I guess they think they will be able to rent it nonetheless. The previous project would have added 800 residential units, but now it's only going to be 120. Some people fear new buildings casting shadows, but as the world gets hotter, we need to recognize that building shadows are a benefit to pedestrians, not a cost, here's an NYTimes opinion arguing just that. I'm jealous of downtown Toronto, where they're about to put up three towers this tall with over 2300 residential units between them.
Here's the full article "AI as Normal Technology", I've heard the story of electricity coming into factories slowly and only after massive retooling multiple times, but it's a good reminder that the template for huge transformations is still relatively slow progress.
Apparently the drone’s details are still secret too, presumably it’s reusable somehow. Well done Eric Schmidt
Flotsam and Jetsam
– California's SB79 has gotten a bit weaker as it's wound it's way through the assembly, but I'm still hopeful that it'll pass and give us some upzoning near transit. (src)(cached)
– Ted Cruz personally was responsible for adding language into the OBBB that cut funding for weather forecasting, ironic considering the Texas flooding. (src)(cached)
– Some interesting stats: "Caltrain ridership is surging: Doubling over past 2 years & up 60% over past year. Ridership is now ~2/3 of pre-COVID. It’s rising every month" good to see that electrification is helping things along. (src)(cached)
– In response to an argument that child benefits don't do much because it takes something like $400k per kid, Lyman Stone argues that for a country like Taiwan, that price is clearly worth paying. As another commenter puts it succinctly (cached) "Child benefits raise fertility, you just have to be willing to spend big..." (src)(cached)
– Obama weighs in on the Abundance movement, essentially saying that it doesn't matter how deeply you care about people if you're unwilling to build anything (src)(cached)
– Multi agent AI systems seem sorta stupid, given that all the agent models are stateless, like it's just a bunch of perfectly identical clones handing papers around as if one of the other identical agents would somehow do anything different. But this commenter had a useful way of explaining what the point is: it's "purely a context management hack, a way to get sub-tasks done without burning a load of your precious token allowance". And that makes sense to me. (src)(cached)
– Trump spent some time signaling that he's unhappy with Fed chair Powell and might try to fire him and appoint someone new. Some of the regional fed presidents have made it clear that if he does that, they'll just ignore the new appointee. (src)(cached)
– Apparently some people are under the impression that the Trump admin shut down PEPFAR accidentally and that it was only temporary, but no, they've continued to work hard to shut it down. (src)(cached)
– User TracingWoodgrains got into a discussion about how in modern social media, some dogwhistles are actually powerful because not only do they signal alignment with one group, they also can cause their opponents to react extremely negatively to something that seems to bystanders to be benign. You have to know when to hold back and not yell that something is actually meant to signal alignment with fascism or whatever. But there's definitely a point at which yes, you do need to point and say "yo, that's some nazi stuff right there". (src)(cached)
– OpenAI has claimed that one of their unreleased models got a score qualifying as a Gold medal in this year's International Math Olympiad. Terence Tao, possibly the smartest mathematician alive, has some commentary on the subject. Essentially, he's saying that OpenAI hasn't been clear enough about what they're saying they've accomplished. (src)(cached)