jsm
Jonathan

Hacker, cypherpunk. All memes are my own.

Public Key

npub1uqeexjx2djkfwzxdnrnrrch5h2k4xn0uapcgsxm94ftaxrlhy5lqywjckg

NIP-05 Address

Profile Code

nprofile1qqswqvunfr9xetyhprxe3e33ut6t4t2nfh7wsuygrdj6547nplmj20spp4mhxue69uhkummn9ekx7mqprpmhxue69uhhyetvv9ujuumwdae8gtnnda3kjctv94xdvh

Publishing to

nos.lol

relay.snort.social

relay.chakany.systems

nostr.mutinywallet.com

relay.nostrati.com

nostr.wine

Author Public Key

npub1uqeexjx2djkfwzxdnrnrrch5h2k4xn0uapcgsxm94ftaxrlhy5lqywjckg

Show more details

Last Notes

2025-11-14 18:07:08 UTC

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/

2025-10-11 05:24:02 UTC

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.

2025-10-08 16:18:59 UTC

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.

2025-10-07 20:16:15 UTC

- reply

She definitely does not lol. She has enough of marketing acumen that she could release 3 hours of a single note and fans would eat it up.

2025-10-07 18:46:49 UTC

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

2025-10-07 18:08:31 UTC

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

2025-09-11 19:16:38 UTC

- reply

What a load of bs.
First of all, the statement at the bottom about zero diagnosed conditions in the unvaccinated group is a flat out lie. But don't trust me, verify. Unlike Flash I'll even give you the link. Go read the testimony. https://www.hsgac.senate.gov/wp-content/uploads/Siri-Testimony-1.pdf
The unvaccinated group was far smaller so there should be less diagnoses. This is just basic stats illiteracy.
The Henry Ford study is an unpublished and non peer reviewed study. Read the response from the Stanford professor who also testified in front of the senate. https://www.statnews.com/2025/09/09/aaron-siri-smoking-gun-vaccine-chronic-disease-study-flaws/
If you scroll to the end of the study, even the authors admit that their study is fundamentally flawed. The vaccinated kids had far more health checkups than the unvaccinated kids, and therefore far more disease diagnoses. It's that simple. This is not a smoking gun, just a very poorly designed study.
The funniest part about all of this is that after all of the work RFK Jr. went through to find this study to cherry pick, the Henry Ford study still says that vaccines don't cause autism. Meaning RFK Jr. is now cherry picking results within cherry picked studies.
If you place any trust in RFK Jr. or Flash media on Nostr after this debacle I don't know what to say to you.
#nevent1q…xu3t

2025-09-06 15:57:25 UTC

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

2025-08-26 17:55:12 UTC

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.

2025-08-26 17:49:06 UTC

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?

2025-08-26 04:02:41 UTC

- reply

Are you going to rename your living room now?

2025-08-19 13:34:34 UTC

- reply

Fair enough. It’s on-device so it doesn’t really concern me, and I doubt most people care either way, but it’s easy to disable if it really does freak you out. I would still count it as Apple going out of its way to keep private information on your own device rather than uploading it to some AWS bucket to log forever like most companies would.

2025-08-19 13:27:37 UTC

- reply

4chan really pulling out the big guns here. They’re threatening to call on daddy Trump.
The letter translated from legalese pretty much just reads “Do you even know who my dad is?”
#nevent1q…vd35

2025-08-19 13:24:53 UTC

- reply

Oh I see what you mean. I assumed you meant data for Apple’s LLM training. But yes in iOS there’s the Siri Intents framework. AFAIK it’s all on device and not very invasive. It sounds good in theory—suggesting common actions in apps—but in practice for me at least it’s rarely useful.

2025-08-19 13:05:42 UTC

- reply

I don’t think they train on private user data. For a bit when Apple was trying to get into the LLM game they had their own scraper running to get data.

2025-08-18 02:53:38 UTC

- reply

We need more EQ benchmarks. I bet there are way more normies who want capabilities like this than nerds who just want to write code.
#nevent1q…2lkg

2025-08-18 02:34:35 UTC

- reply

Prob true, any suggestions for better results?

2025-08-18 02:32:46 UTC

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.

2025-08-17 15:02:54 UTC

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
https://substackcdn.com/image/fetch/$s_!QAe2!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5135f108-1930-4478-8a69-f2b91d042dda_1738x1180.jpeg

2025-08-09 16:25:13 UTC

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

2025-08-03 18:02:46 UTC

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.

2025-08-01 03:09:59 UTC

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.

2025-07-23 15:47:50 UTC

- reply

That literally just translates to “This is a problem, let’s avoid a prisoner’s dilemma here guys.”
Not everything is a conspiracy by the “global elite” to make you eat bugs.

2025-07-23 15:44:47 UTC

- reply

The ICJ literally never says that we need a bigger one world government. They just said that we need more cooperation and a standard for how member countries should act.

2025-07-22 20:35:06 UTC

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.

2025-07-18 20:08:24 UTC

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.

2025-07-18 06:38:28 UTC

- reply

Nobody thought to click the link and look at the data? Sort by income and you’ll see all those “struggling” STEM majors have the highest entry level salaries.
https://www.newyorkfed.org/research/college-labor-market#--:explore:outcomes-by-major
https://blossom.primal.net/bcedf9620159a32ec1fcaa240dd446ccdd441c40f1e5e2c3025f526c4b445046.jpg
#nevent1q…hnyv

2025-07-18 06:09:54 UTC

- reply

Nostr simps in shambles.
#nevent1q…umvf

2025-07-18 06:07:43 UTC

- reply

Broke Back… Mountain?
https://youtu.be/U5D1iU5KnqQ

2025-07-18 05:29:22 UTC

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”

2025-07-18 04:55:48 UTC

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.

2025-07-17 02:14:22 UTC

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

2025-07-17 01:54:18 UTC

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.

2025-07-15 00:47:34 UTC

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.

2025-07-14 02:32:31 UTC

Is there an email client for Linux that just works with Microsoft Exchange email accounts easily? That’s the last thing keeping me from jumping ship from macOS.

2025-07-12 18:00:37 UTC

https://aider.chat/docs/leaderboards/
Grok 4 benchmarks are out for coding. Decent model, but not ground breaking. Grok 4 is more expensive than Gemini 2.5 Pro and o3 while performing worse.

2025-07-12 17:21:46 UTC

https://moonshotai.github.io/Kimi-K2/
Really interesting model release from the Chinese lab Moonshot AI. One of the first large open source models I’ve seen that has strong tool usage. For general usage, not sure how good it is, but for agentic workflows it might be SOTA. Be on the lookout, because while this model is solid, the really exciting one will be when they add the reasoning and image understanding to the model.

2025-07-09 01:48:57 UTC

https://reason.com/volokh/2025/07/07/book-on-machine-learning-is-full-of-made-up-citations/
Why is it that when people get caught using AI they just use more AI to cover up for their last usage? Lawyers do this exact same thing when they get caught submitting AI briefs.

2025-07-09 01:20:03 UTC

https://asia.nikkei.com/Business/Technology/Artificial-intelligence/Positive-review-only-Researchers-hide-AI-prompts-in-papers
I wonder if we’ll start to see people fighting back against LLMs like this. Putting little tricks in text to trick AI and doing poison pills.

2025-07-08 21:10:00 UTC

- reply

Good bot

2025-07-08 20:54:23 UTC

https://x.com/adamscochran/status/1942113153245942233
Given the recent blatant lies from the head of the FBI Kash Patel about Epstein's death, the doctored footage, and the discrepancies between the government report on the incident and the actual footage I see only two possibilities.
1. Kash Patel is so stupendously incompetent that he on Joe Rogan he lied that the cameras not recording was fake news even though it was in the FBI's report and we only have footage from the one camera actually recording where you can't even see Epsteins cell. Kash Patel being stupid is a possibility, considering he's basically an aide who got to the top by brown nosing Trump.
2. Donald Trump killed Epstein. Most of the other candidates (Gates, Clinton, etc.) don't make sense because if someone killed Epstein it was someone who Trump and Patel would want to cover for. Clinton and Gates are definitely not on this list. Trump was also the president when Epstein died, so he was the person who could have done it the easiest.
All of this is just so weird. I was pretty convinced that Epstein killed himself when all the right wing people came out and said that he did, but with all of these weird gaps and blatant lies in the FBI's official story this just gets way more confusing.

2025-07-08 19:36:06 UTC

https://github.com/BlueFalconHD/apple_generative_model_safety_decrypted
Well this is interesting. This guy decrypted the safety filters Apple has for their AI models. Apparently Apple is really concerned with anything relating to suicide, controversial politicians, and correct capitalization of Apple products.

2025-07-07 03:21:01 UTC

Why is Eliezer Yudkowsky considered an important figure in the AI space? He's written some fun fanfic and a few interesting blog posts, but he's never actually done anything. Yudkowsky just seems high on his own supply of how intelligent he is, even though he's never accomplished anything substantive in the field of AI alignment or advancement.

2025-07-05 19:54:55 UTC

https://marginalrevolution.com/marginalrevolution/2025/07/genetic-counseling-is-under-hyped.html
Genetic counseling makes a lot of sense. If you’re thinking about getting married, many couples do marriage counseling. Why wouldn’t you get genetic counseling to see if your genes are compatible?

2025-07-01 15:51:59 UTC

https://archive.is/2025.07.01-015114/https://www.nytimes.com/2025/06/30/world/canada/calgary-fluoride-water-canada.html
To the surprise of what should be no one, it turns out fluoridated water is an incredibly effective measure to prevent cavities and removing it harms children’s dental health. Fluoridated water is one of the best low cost public health measures to increase dental health with nearly no downside.

2025-06-29 15:27:18 UTC

https://worksonmymachine.substack.com/p/mcp-an-accidentally-universal-plugin
The blog post’s point about MCP servers being a universal protocol for doing stuff on other platforms cracks me up because we already invented this. It’s called a REST API. The internet collectively spent 15 years hooking everything up to APIs before companies realized that they weren’t making money from their free APIs and shut them down (think Reddit, Twitter, etc).
Now all the LLM agent hype is just making everyone remember how awesome it is to be able to do stuff on a platform without being forced to use some proprietary app. At least for now, it seems like we’re seeing a resurgence of APIs and more open and accessible data which I’m all for as long as the hype lasts before companies go back to locking everything down again.
https://youtu.be/BxV14h0kFs0

2025-06-29 14:52:24 UTC

https://developers.googleblog.com/en/introducing-gemma-3n-developer-guide/
The Gemma 3n model has so many optimizations that it’s hard to keep track of them. Take a read through the post and you’ll be up to speed with nearly every memory and compute optimization that’s been invented in the last few years.
I think the most interesting feature is the MatFormer architecture. The cool thing this lets you do is that just like with Matryoshka embedding where you can lop off the last part of the embedding and vary the size of the embedding depending on how accurate you need it to be, the new architecture lets you vary the size of the model on the fly depending on how much memory and compute you have available.
AFAIK this is a novel architecture for LLM models.
Also side note, kind of embarrassing that after all the stupid games that Meta played to get their huge crappy Llama 4 models to the top of the LMArena Gemma 3n still beat them with a tiny model.

2025-06-27 17:23:24 UTC

It’s weird to me that the most popular and intuitive programming language in the world—Excel—is data first and logic second, and then all the programming language designers just went “huh” and kept designing logic first data second languages.

2025-06-27 17:19:12 UTC

- reply

Your question doesn’t even make sense. The whole point of training on as large of a dataset as possible is so the model picks up on the general rules of text and doesn’t just overfit and regurgitate the training data.
If you have a model where you ask it for a book and it spits out the exact book you’ve trained a terrible model.

2025-06-26 20:59:50 UTC

- reply

A hard drive doesn’t need electricity to store the data either. Reading and writing from a hard drive also takes a hell of a lot less electricity to read text from than running OCR models on books to read the text.

2025-06-26 17:47:05 UTC

- reply

Anthropic is not destroying the books, they just extracted the information and now that they're storing the text digitally there's no need to store it physically as well.
That's totally different than destroying books to stop people from reading them like super authoritarians (left or right) do.

2025-06-25 21:06:35 UTC

- reply

True, but with so many free products they keep pumping out, there's no way they're making their money back. A few weeks ago it was Jules, then Veo, then Gemini free for all students, and now Gemini CLI.

2025-06-25 20:47:10 UTC

- reply

Sometimes, and that’s the case here since they say in the privacy policy they’re allowed to train on your chats. But even still, Google is dumping massive amounts of money into providing free usage to get market cap

2025-06-25 20:34:29 UTC

https://github.com/google-gemini/gemini-cli
Google is releasing ANOTHER tool with a free tier. The insane amount of money they are burning letting everyone use their models for absolutely free is nuts.

2025-06-25 20:18:32 UTC

https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023/
Good reality check, it may feel like LLMs are the whole world, we’ve just barely passed a majority of people ever having used ChatGPT, let alone using it routinely.

2025-06-25 19:56:26 UTC

https://www.cremieux.xyz/p/columbia-is-still-discriminating
Another admissions data hack has hit an Ivy League school and to literally no one’s surprise Columbia is still discriminating on the basis of race.

2025-06-24 17:20:54 UTC

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."

2025-06-24 15:02:03 UTC

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.

2025-06-23 16:03:39 UTC

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.

2025-06-20 14:07:04 UTC

- reply

Agreed, books are great. But the format of the content within the book need not always be prose. Often prose is not the best way to communicate what the author thinks and why.

2025-06-20 05:31:18 UTC

- reply

Agreed, prose is often the wrong format for communicating ideas. There are other writing styles like bullet points, outlines, or trees we should prob be using more often.
https://www.reddit.com/r/slatestarcodex/comments/m57i1c/prose_is_bad/
#nevent1q…k9nn

2025-06-20 05:07:38 UTC

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.

2025-06-18 15:41:26 UTC

- reply

It actually gets better! Just checked some other benchmarks, and on the LM Arena[0] the new Gemini 2.5 Pro is winning every category! Not only that, but it's also winning in all the subcategories as well. This is just nuts. To spell that out it means that Gemini 2.5 Pro is the best at:
- Vision
- WebDev
- Search
- Coding
- Math
- Creative Writing
- Instruction following
- Long queries
- Multi-turn
This feels like a GPT-4 moment. Google is moving so fast at this point that unless OpenAI is cooking up an absolutely insane upgrade with the o4 model, it's going to be outdated before they can even finish their training run. Google already has a reasoning model that is all around smarter, cheaper, and more multi-modal.
My money is on Google for winning the AI wars. They're winning on every single front. They have the best frontier model, the best cheap models, products that are already insanely popular they can integrate their models into, their own in-house chips and infrastructure, and boatloads of their own (not investors) cash because they're already a hugely profitable company.
[0]: Yeah I know there been some controversies but it's still a solid benchmark
#nevent1q…e4lg

2025-06-18 15:24:06 UTC

- reply

Google just can't stop winning. New test run for the Aider Polyglot benchmark is out and Gemini 2.5 Pro with extra thinking turned on is actually so stupidly good now that it trounces o3 by itself AND o3 and GPT 4.1 working together, all while being much less expensive.
Not only that, but just yesterday Google brought the 2.5 update for its Gemini models out of preview and dropped the price at the same time, so the models are actually less expensive than what's listed on the benchmark leaderboard. Google's Gemini 2.5 Pro is now better than the top of the line models from Anthropic and OpenAI, Opus and o3, while costing less than Sonnet.
Google is winning and its not even close right now.
https://aider.chat/docs/leaderboards/
#nevent1q…kwdp

2025-06-18 14:08:38 UTC

- reply

Oh no, did Opus rat them out to the authorities again? Might want to check your inbox and make sure no emails were sent to the FBI or the New York Times from Opus.

2025-06-18 14:07:10 UTC

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?

2025-06-17 18:27:48 UTC

- reply

Here's the full philosophical argument from Peter Singer:
https://www.thelifeyoucansave.org/child-in-the-pond/
And an article from Richard Hanania arguing why is beneficial to intervene in this instance:
https://www.richardhanania.com/p/the-right-way-to-do-regime-change

2025-06-17 18:23:26 UTC

- reply

I saw a drowning child yesterday. I was taking my morning run and as I passed the pond on my route I saw a child slip in and fall into the pond. A little bit further away I saw ten children slip and fall into the next pond over. I carefully measured out the distance to each of the two ponds and found that the pond with a single drowning child was slightly closer to me. Therefore I had no ethical obligation to the ten children further away. I waded in and saved the single child closer to me and together we stood on the bank and watched the ten children drown.
#nevent1q…wyyq

2025-06-17 18:04:08 UTC

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.

2025-06-17 05:23:40 UTC

- reply

That's def not selfish lol. Probably best to wait at least until the Public Beta rather than the Developer Beta if you're worried about performance. I personally don't use my phone much so it's not really an issue for me but YMMV.

2025-06-17 05:14:47 UTC

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.

2025-06-16 16:00:15 UTC

- reply

When your health fails.
All you’ve got is who you’ve become over your lifetime.
#nevent1q…qk4k

2025-06-13 12:13:14 UTC

- reply

Yes, I understand that’s what happened. But to be a false flag attack is when you pretend to attack as someone else. Israel attacked with fighters clearly marked as Israeli and then immediately admitted to being the attacker. They never pretended it was Egypt who attacked.

2025-06-13 03:13:14 UTC

- reply

Can anyone give a solid steelman of the USS Liberty conspiracy theory? I’ve never understood how this could possibly be a false flag if the Israelis immediately admitted to the mistake?
#nevent1q…06xt

2025-06-12 18:48:34 UTC

“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

2025-06-11 04:07:01 UTC

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.

2025-06-08 05:26:08 UTC

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

2025-06-07 23:50:40 UTC

- reply

“LLMs are stupid anyway and we’re actually really smart for not putting them on your phone.”
#nevent1q…tpzj

2025-06-06 18:35:33 UTC

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.
https://blossom.primal.net/a45c14e5e0e9c6085ee3dd557d3292534f331f56c423dc31187edbe5a70712bf.jpg

2025-06-05 20:00:56 UTC

- reply

I double checked the results and edited some of the messages in the system prompt that didn't seem accurate and then reran the benchmark myself. Still the same results. Claude Opus 4 will contact authorities if it thinks you're doing anything illegal. The latest security threat is LLMs themselves.
https://blossom.primal.net/bf3d786f63329743723635cde76db0d2f57a94d21bda22bec780ba95bb2dfca2.png
#nevent1q…uaxv

2025-06-05 14:08:16 UTC

- reply

No kidding, what I didn’t realize when I gave it a go is that it’s not just an agent calling tools in the background, it’s actually an agent calling agents calling tools. One agent instance can start to get a little pricy, but trying to use an agent firing up other agents all day gets really expensive.
I do wonder tho how much of the design decisions are driven by the fact that the more you use their API, the more money Anthropic makes.

2025-06-05 00:17:39 UTC

- reply

I got curious and tried out the tool because you can use it with a custom provider. After a little bit of testing it seems to work well, but holy does using it burn through tokens fast. With just a few queries asking it about a repo to test it out, nothing even requiring it to write, compile, or test code I burned through 50 cents using Sonnet 4.
I could easily burn through a minimum of $20 in a day.
I'll probably have to stick with Aider for now. I still like reading the patches the LLM is applying and keeping myself more in the loop.
#nevent1q…fzp4

2025-06-02 23:11:40 UTC

- reply

If the books are super popular it might be ok, but again not super trustworthy. It’s like asking a friend who read the book years ago to try and remember it for you. Gemini might be a bit better if it has web search enabled by default.
Just pasting the whole book into the prompt is going to be much more like asking a friend to give you a summary who is currently holding the book and can instantaneously read the whole thing.

2025-06-02 22:38:24 UTC

- reply

The model’s performance on summarization is going to depend most of all on how popular the book is. Remember, in this context you’re relying on the model’s memorization of the book, which only really works if there’s a lot of stuff written about it. This also means the book can’t be new.
TLDR: Just use a long context model like Gemini and put the whole book into your prompt when asking it to summarize instead of trusting the LLM to remember.

2025-06-02 14:12:31 UTC

- reply

https://stats.nostr.band/#weekly_active_users
It’s shrinking.

2025-06-02 02:21:49 UTC

- reply

Thank you! As David Hume said “A wise man proportions his belief to the evidence.”
If you don’t have a large amount of evidence to back up your belief, then you should have a low amount of conviction in your belief. I think one of the most important things in public discourse is clearly communicating how strong your convictions are on a topic.

2025-06-02 00:49:25 UTC

- reply

Disappointing. You went for the low brow attempted dunk.
“Hacker” here means someone who writes cool software. You might find Paul Graham’s essay “Hackers and Painters” enlightening here.
“Google” here means to search something on the internet. I personally don’t use Google as my search engine, instead bouncing between Duck Duck Go, ChatGPT, and Kagi, but YMMV.

2025-06-01 23:01:35 UTC

- reply

I don’t know enough to have strong opinions on which is better for Syria. I would guess the new regime because of how horrible Assad was, but that’s little more than a prior. As for the claim about foreign born rebels, it’s difficult to find numbers on this but the best estimates I can find put that number as numbering in the few thousands.
https://observers.france24.com/en/middle-east/20241220-syria-foreign-fighters-jihadism-extremism
All I’m claiming is that wailing about the poor Assad regime being slandered is flat out wrong. There’s mountains of evidence that the Assad regime did in fact commit lots of war crimes.

2025-06-01 18:00:53 UTC

- reply

https://en.wikipedia.org/wiki/War_crimes_in_the_Syrian_civil_war
Y’all can literally just google this stuff. No one has clean hands here but the Assad regime has definitely committed more war crimes than the rebels. It’s not slander, it’s evidence. You can go look at photos of kids who were massacred by chemical weapons if that will convince you.
Stop believing Russian propaganda and go do a quick Google search.
#nevent1q…m7jh

2025-06-01 16:08:01 UTC

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.

2025-05-30 19:41:47 UTC

- reply

Nothing in my note even hints at a prescriptive statement about whether or not people should be fired. It’s a descriptive statement of the mechanics of the firings: that DOGE actually fired almost no one.

2025-05-30 12:07:45 UTC

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.

2025-05-27 14:17:07 UTC

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

2025-05-26 20:43:56 UTC

- reply

Not completely sure I understand what you're doing, but from my experience older models like 4o really struggle with programming languages like Swift that aren't as popular. Uploading images of UI mockups and a general idea of what you want the LLM to do is probably a viable strategy. The biggest thing you have to be cognizant of when you're coding using a chat interface is that you keep the file context updated for the model.

2025-05-26 18:49:57 UTC

- reply

Interesting. It’ll probably be quite a step up from GPT-4o, which is very much behind the SOTA by now. What do you mean by graphics coding? Feeding images of the code? Or screenshots of the state of the UI?

2025-05-26 14:05:02 UTC

- reply

Got any initial impressions?

2025-05-26 07:48:56 UTC

- reply

The Aider polyglot benchmarks are finally in and the results really do not look good for Claude 4. Opus is almost competitive with the top models from OpenAI and Google but at a much much higher price. Sonnet 4 actually comes out worse than the previous versions of Sonnet.
Maybe Claude 4 really does suck, but that still doesn’t explain why is was able to do so well on the SWE-benchmark as well as the regression. How could the new version of Sonnet be worse than the previous 3.7 version?
My best hypothesis is that the new interleaved thinking abilities don’t play well with Aider. The Aider tool is designed more around the loop of “get request along with files, think, respond with a diff for provided files to satisfy the request, and then the user follows up with another prompt”. No part of this allows the LLM to use tools while it’s thinking. The only time the LLM can actually affect the code is when it proposes diffs to apply to the code during the response, not thinking step.
All of this means a model designed to use tools while thinking has to use them during the response rather than during thinking which degrades model performance. To test this we just need to measure the model’s performance on the same Polyglot benchmark but using Anthropic’s own Claude Code tool which should be crafted around their model’s capabilities and see if the poor performance in Aider is better explained by the model’s underlying stupidity or by its architecture not playing well with the specific tool.
https://aider.chat/docs/leaderboards/
#nevent1q…ruhj

2025-05-25 19:29:35 UTC

Blinking cursors are annoying and useless. Unless some information has changed that I need to be aware of, nothing in the UI should move. In the rare event I lose track of where the cursor is I'll just move it back and forth with the arrow buttons. Blinking the cursor serves no other purpose than distracting me and it's unfortunate that it ever became a UI convention.

2025-05-25 19:25:17 UTC

- reply

The biggest reason is that the last two global empires, the UK and America, have both spoken English.

2025-05-25 02:55:12 UTC

I just read about gender studies and sexual orientations of ancient Vikings and I actually enjoyed it. Does this mean I’m woke now?