Jonathan

I’m curious why there’s a drop in the number of girls wanting to not get married but boys have stayed basically constant. The survey definitely shows like women have a lower desire for children, but how much does the decline in the institution of marriage play into this rather than an underlying change in a desire for families? https://www.pewresearch.org/short-reads/2025/11/14/12th-grade-girls-are-less-likely-than-boys-to-say-they-want-to-get-married-someday/

https://pwascore.com/ I’ve never realized what an ungodly hulking behemoth the modern web browser has become until reading through all the features required for PWAs.

Updated to macOS Tahoe. Of course Apple broke a whole bunch of scripting APIs that I use. Apparently you just can't get the currently playing song title in Apple Music with Applescript because fuck you I guess. Every update, I start to more seriously consider switching to Linux for my next laptop. I hate feeling like I'm a guest on my own computer begging for the ability to get access to the internal state of my own laptop.

The new Taylor Swift album feels like the songwriter version of computer scientists whiteboard masturbating. She spent so much time flaunting how clever she is at writing lyrics that she forgot to actually make good music. Everything just comes out sounding blandly floaty. https://www.urbandictionary.com/define.php?term=whiteboard%20masturbation

Just realized that all the icons in the macOS System Settings app change sizes when the window is frontmost vs not. Why in the world would this be helpful? https://blossom.primal.net/a3cfb9885692158dc29a67735cd819d11b1a62de3e97867e41843cf9a4f2bc34.mov

https://tempo.xyz/ It’s 2021 and people are launching stablecoins again.

https://manifold.markets/nathanwei/will-the-un-declare-a-famine-in-gaz Can't believe I got rug pulled by the UN. What a joke. I forecasted that the UN wouldn't declare a famine in Gaza because a real famine requires 2 deaths per 10k, which works out to ~420 deaths per day. The chances of this happening are insanely low and you've been brain rotted by your news feed if you think it is. What I failed to factor in was the possibility that the UN would just change its definition of what a famine is by removing the 2 deaths 10k requirement completely. The situation in Gaza is definitely bad, but focusing this much on what words we use to describe it (genocide, famine) is not helpful and distracts from the actual situation on the ground. Semantics are stupid, just focus on the facts.

Kinda can't believe Taylor is actually getting married. What will all those celebrity shows do now that they can't speculate on her love life?

Note to self: never ask an LLM for flirting suggestions. It'll take a while for capabilities to improve enough to become useful. Just wrote a Python script to extract messages from a conversation in iMessages, then fed it into an LLM to get feedback and suggestions as a test. After reading LLM flirting suggestions I think I might die of cringe. I have never felt such a strong urge to gag while reading non-NSFW text.

For all the other people who may have, like me, seen this meme graph all over the internet with everyone attributing it to their favorite pet theory for the decline of society, it turns out the data had to be tortured pretty hard to get this. https://grimoiremanor.substack.com/p/no-conscientiousness-hasnt-collaped

I finally made a decision and took the Giving What We Can pledge. Instead of feeling helpless about the evil going on in the world, do something about it and improve the world as much as you can with what you have. Don't settle for giving to the charity that has the best marketing materials or the most emotional appeal, look at the numbers and give to the charities who are doing the most good for the least amount of money. Every dollar matters. Every person can easily save many lives every year. Make a difference and take the pledge. https://www.givingwhatwecan.org/ "Saruman believes that it is only great power that can hold evil in check. But that is not what I have found. I have found it is the small things, everyday deeds of ordinary folk, that keeps the darkness at bay. Simple acts of kindness and love." - Gandalf

https://leavesubstack.com/ If you want to leave Substack because you don’t want to be beholden to a centralized platform, then great! That’s a decent reason to leave. But everyone whining about Substack not having tight censorship is silly. The only reason there’s censorship online is because of advertisers. Companies don’t want their names next to Nazi propaganda, which is fair. Substack is subscription based so there’s no point in all the pomp and show of censorship. Censorship is stupid, leaving Substack because you want Substack to decide what people can and cannot write is stupid.

https://marginalrevolution.com/marginalrevolution/2025/07/say-it-aint-so-cecil.html I when walking the fine line between safety and privacy, I tend to come down on the side of privacy, but it does seem strange that breathalyzers aren’t required on cars. They’re not very privacy invasive and it seems like it would do a lot of good.

https://towardsdatascience.com/text-generation-gpt-2-lstm-markov-chain-9ea371820e1e/ Great reminder of how much goal posts change. The author of this article is showing a survey of different text generation methods in 2021 before ChatGPT and is impressed that GPT-2 can not only write mostly grammatically correct and logical sentences, but the model can use correct pronouns to refer to people ("her" when referring to "sister"). We've come a long way.

One of the best things you can do when you read something is to simply click links. That’s it. When an article makes a statement and has a link, just go ahead and click it and then start reading. Forcing myself to engage and research what I’m reading rather than passively consuming has had a lot of benefits. First, it lets you prune out people or sources you read that may seem trustworthy on the surface but are citing absolute crap to support claims. Second, reading more primary sources inoculates against a lot of lazy writing about “scientists say X”. You’ll quickly realize there’s a broad range of research credibility and rigor. Just because there’s a study that says something that doesn’t mean it’s true. Third, it’s just a good brain exercise. Force yourself to do hard things. Push back against the TikTok and social media induced stupor of consumption. Reading dense primary sources forces you to put some effort into comprehending what you’re reading. Too much content online is written to require as little effort as possible. Put some mental reps in and get effortful reading back into your daily routine.

https://archive.is/2025.07.17-225639/https://www.wsj.com/politics/trump-jeffrey-epstein-birthday-letter-we-have-certain-things-in-common-f918d796 Is the whole Epstein fiasco what is finally going to be the straw that breaks the camel’s back for Trump’s cult of personality? The whole thing just continues to get more strange and muddled. According to the WSJ, Trump (along with a few of Epstein’s other close friends) wrote Epstein a bawdy letter on his birthday where he talks about their “shared secret”

https://marginalrevolution.com/marginalrevolution/2025/07/the-political-culture-holiday-culture-that-is-french.html It cracks me up that just like we argue over cuts to different government programs like entitlements or foreign aid you could also debate federal holiday cuts.

https://thehill.com/homenews/senate/5402273-white-house-accepts-pepfar-exemption/ More good news! The best parts of USAID are being preserved. PEPFAR has been an incredibly cost effective program that’s saved over 26 million lives from AIDS. https://www.state.gov/results-and-impact-pepfar

https://mozillagfx.wordpress.com/2025/07/15/shipping-webgpu-on-windows-in-firefox-141/ Finally getting WebGPU support in Firefox! Lets gooo! There’s all sorts of interesting applications you can build that require GPU access to be remotely performant that have been handicapped by Firefox being incredibly slow to support WebGPU.

https://kiro.dev/blog/introducing-kiro/ Days without new AI IDE: 0 Interesting concept tho. The basic idea is to break every task down into smaller and smaller subtasks to implement. This might be a more scalable architecture, since you can assign agents to different tasks simultaneously. It also probably scales better with AI capabilities since the AI can break each task down into fewer larger subtasks as model capabilities increase.