Jonathan

I've been living for two days in the overlapping area in a Venn diagram of the absolute worst kinds of debugging: network firewall config and Python conda dependency hell. With all the new Apple Health mental health stuff they should add a feature that's like "It looks like this is the third time you've tried to prune the conda base environment, and you're starting to swear in your messages to the LLM. I'm phoning the suicide hotline now."

The whole pre-occupation with consciousness when people talk about AI seems confusing to me. It feels like people talk about whether AI is conscious like it’s an important or meaningful quality to have without ever explaining why consciousness is important in the first place. My personal guess is that consciousness is a psychological quirk of how our minds evolved. I don’t see anything particularly necessary about it. Perhaps a conscious intelligence is useful in some contexts, but it seems possible to have non-conscious intelligence that are just as, or more, powerful than their conscious counterparts.

After trying to cook with Beef Tallow, I now understand why we’ve moved away from cooking with it. Tallow’s smoke point seems really low and I had to open all the windows to avoid setting off the smoke alarms. It also splattered out of the pan all over the kitchen. Health theories aside, I understand the move to seed oils much more. Not only are they cheaper, they’re just better to cook with.

The Dow Jones doesn’t make sense. It’s price weighted, which doesn’t make sense considering a tiny company with a few expensive shares can have more influence than a massive company with lower priced shares. The reason is that the Dow was started in 1896, a time when obviously computers didn’t exist. Getting the data and crunching it every night was a huge computing task, so they had to settle for the price since that was quicker to calculate than multiplying out the total market cap. It’s strange to think that there was a time where the basic task of multiplying a list of numbers was an expensive and time consuming process that required paying a team of people. The assumption that every single person has instant access to a computer capable of processing billions of computations of massive numbers every second is so ingrained into my brain that it’s strange to imagine a time when that didn’t exist.

My new favorite LLM bug that I've seen is that sometimes when the context gets too long with DeepSeek R1 the model starts spitting out random pieces of training data in Japanese. It'll be in the middle of generating a diff for for some code and it just starts spitting out Kanji in the middle of it. My favorite part is that the model itself seems super confused and during its thinking process says stuff like "oh my it looks like I didn't generate that diff correctly, let's try that again and do exactly what's in the file" I (unsurprisingly) can't read Kanji but I tried Googling some of the stuff and it seems like the little inserted bits of text are excerpts from Japanese government or university websites?

https://entropicthoughts.com/intuiting-monty-hall Actually intuitive explanation of the Monty Hall problem. Most people who know the answer to the puzzle know precisely that: just the answer, but it’s famously hard to truly grok the problem.

I installed the new iOS 26 update on my iPhone. Obv as a dev beta right now so the performance is terrible and there are tons of bugs. The new "Liquid Glass" paradigm feels, eh. It doesn't feel modern and streamlined like Big Sur, more like one guy on the iOS team figured out how shaders work and applied his favorite opacity/diffusion shader effect to everything. It's an interesting look for the first few hours until the novelty wears off. Not about accessibility tho. I have young eyes so readability wasn't an issue but I'm not sure if older people will be able to say the same.

“Inspired by a mind-expanding LSD journey in 1985, I designed the HyperCard authoring system that enabled non-programmers to make their own interactive media.” Full article is fascinating, but this part in particular stuck out to me. I’d never considered how much of a role psychedelics might play in software design. https://www.folklore.org/Joining_Apple_Computer.html

https://simonwillison.net/2025/Jun/10/o3-price-drop/ OpenAI was able to optimize their inference for o3 by 80%. This is nuts. I wonder what they were doing before that was so inefficient. It’s now cheaper than Claude 4 Sonnet and Opus.

Gives you perspective. https://lnfly.albylabs.com/api/apps/296/view?npub=npub1uqeexjx2djkfwzxdnrnrrch5h2k4xn0uapcgsxm94ftaxrlhy5lqywjckg

Watching the fallout on the right from the Trump-Musk fight is hilarious. We need to open some betting markets on who flips which way.

Everyone is using AI, at every level of society. The MAHA report released by RFK Jr. was written by ChatGPT. WaPo compiled the pretty damning evidence. The funniest part is that everyone has been arguing about how or if AI will somehow manipulate leaders to gain control of power. No one considered that leaders would just ask ChatGPT to take power. https://web.archive.org/web/20250601160757/https://www.washingtonpost.com/health/2025/05/30/maha-report-ai-white-house/

https://simonwillison.net/2025/May/31/snitchbench-with-llm/ What is Anthropic training Claude Opus 4 on? First the system card said that if you try to shut it off and the model has access to potentially embarrassing information (like an affair you’re having) it will attempt to blackmail you. Now new tests are showing that if the Opus model finds anything it deems morally objectionable in your email or logs it will take it upon itself to contact government authorities or the media to rat you out.

https://sahillavingia.com/doge This article has changed my understanding of DOGE. Apparently a lot of the firings weren’t done by DOGE, because they don’t have any real power, the agency heads did those all on their own and let DOGE take the fall for it. DOGE is more of the scapegoat for the other departments.

https://www.noahpinion.blog/p/the-age-of-twitter-is-finally-ending Twitter is dying. What social media platform will take its place?