Played around with image generation using Flux in ComfyUI and a couple of LORAs. Decided to update my profile image and banner. I generated my old ones a long time ago when DALLE-2 first came out.
Now thinking of using Wan 2.1 to turn it into a gif, but I don't think keyframes are available yet with Wan in ComfyUI so it would be difficult to make a perfectly looping gif.
https://livecodebench.github.io/leaderboard.html
Currently the best model by far for coding is Kimi by the Chinese company Moonshot. It's trouncing every American model. Even though it was released weeks ago I've never heard of it. I can't find literally a single website in English mentioning it. This feels like DeepSeek R1 again but no one has any idea.
https://arxiv.org/html/2501.12599v2
Finally found the paper at least. I had to go to the wasteland most have never traveled to: the rumored second page of the search results.
Some interesting techniques that I haven't seen before. They made sure the problems they did the RL on really did requiring reasoning to solve by having a non-reasoning model guess the answer, and remove the problem if the answer was guessable without reasoning. They also modified the reward function to train the model to get the right answer with the least possible amount of thinking. That might help avoid some of those annoying times with DeepSeek R1 and other models where it thinks of the right answer and then says "Wait" and goes off in the completely wrong direction.
Interesting stuff. Still can't figure out why this model release stayed completely under the radar. You would think the Twitter AI influencers would be talking this because Americans are still worried about China having better AI than us.
https://livecodebench.github.io/leaderboard.html
Currently the best model by far for coding is Kimi by the Chinese company Moonshot. It's trouncing every American model. Even though it was released weeks ago I've never heard of it. I can't find literally a single website in English mentioning it. This feels like DeepSeek R1 again but no one has any idea.
I sleep easy at night knowing that if I ever take an incriminating or embarrassing photo no one will ever be able to find it in the thousands upon thousands of memes I’ve saved over the years.
You could also do weekly sponsored posts like lots of blogs. Build and keep an audience of eyeballs with interesting content and then sell that attention to companies or individuals.
Well, I hope this is both right and that the utopia scenario happens.
Or it’s hype and we hit a wall for a while with current approaches. Or we all die.
This is the crème de la crème of 4chan schizo posting. Pull together multiple bad ideas in a way that almost makes sense if you squint to show that it’s all multi dimensional chess in a multiverse. There's just a few problems.
1. Bonds are fixed rate. You can't just refinance them like a mortgage and they're not callable so you can't buy them back early [0].
2. Deflation would indeed lower the nominal prices of goods but it would also gut borrowers (the US gov) [1]. This is paradoxical with paying off the debt.
3. American farmers already have paper thin margins, tariffs on our exports won't mean that farmers will lower prices and start operating as charities [2]. Tariffs also mean higher costs of any imported food.
4. Tanking the stock market does lower the value of the stock people own but it doesn't make poor people any richer. If you (a rich person) buys a share of a company (stock) and that company then is hit hard by tariffs the value of the company will go down (your stock will lose value) and the company may have layoffs so the employees lose to. A shrinking business doesn't benefit anyone.
5. The price of eggs has very little to do with either Trump or Biden's economic policies and much more to do with the recent avian flue outbreak that led to a culling of the hen population [3].
[0]: https://finance.yahoo.com/news/trumps-wacky-plan-for-the-federal-debt-181649828.html
[1]: https://www.forbes.com/advisor/investing/what-is-deflation/
[2]: https://ag.purdue.edu/commercialag/home/resource/2020/10/operating-profit-margin-benchmarks/
[3]: https://www.cdc.gov/bird-flu/situation-summary/data-map-commercial.html
nostr:nevent1qqsfhm5ch4d3e4xj6s2u2fuvdqgkeqej6uuwmrgxe42p9vjwpfxj06gpp4mhxue69uhkummn9ekx7mq7496wc
Why so? They seem like a solid improvement on the original Switch.
Nintendo sent an update. It’s over for Trump. There’s going to be a revolution if he doesn’t repeal the tariffs. /s
https://www.astralcodexten.com/p/introducing-ai-2027
Scott Alexander’s post on impressive AI predictions from 2021. Super impressive stuff considering ChatGPT hadn’t been released yet.
The one major thing that I think Scott is mistaken on in his blog post is his idea that China will “wake up” and steal US weights. The Chinese don’t need to. Alibaba (Qwen), 01 ai (Yi), and DeepSeek are already toe to toe with the best American models. Just as we figured out in 2024 that OpenAI has no moat, we’ve learned in 2025 that neither do American companies.
https://boardgamegeek.com/boardgame/244992/the-mind
Shower thought I had a while ago: This game is trivially easy to solve with sleep sorting.
As long as everyone starts mentally counting at the same time and can keep track of time there’s no luck or telepathy required to sort a list of numbers in ascending order without communication.
https://news.ycombinator.com/item?id=17688725
Looked it up and of course someone on Hacker News already thought of this.
https://boardgamegeek.com/boardgame/244992/the-mind
Shower thought I had a while ago: This game is trivially easy to solve with sleep sorting.
As long as everyone starts mentally counting at the same time and can keep track of time there’s no luck or telepathy required to sort a list of numbers in ascending order without communication.
https://nymag.com/intelligencer/article/how-the-guardian-us-flourishes-without-a-paywall.html
Donating for good content (V4V) is both not at all a new concept and also does work.
https://www.cremieux.xyz/p/fertility-policy-for-rich-countries
Essay arguing for the Hanson Scheme of Social Security where retirement is directly funded by your children. Hanson Scheme alone fails in too many circumstances but perhaps a hybrid model might work?
I can think of a few failure modes for the Hanson Scheme off the top of my head from personal experience. What if you don't have a disability but your children do? Say if your children are born with genetic disorders that lower their life expectancy and therefore their earnings potential. Or set aside genetic disorders completely. How do we deal with good parents who raise their children well into adulthood but then the kids die in an accident? How do you decide what is in the parent's control and what isn't? Is there a way to distinguish between people who go to dangerous places and work for little pay in a situation where it isn't safe to raise a family from the people who don't have a family for more selfish reasons and should therefore use the extra money they got from not having to invest in children for their own retirement?
The author also brings up IVF or adoption as a solution for infertility, but how do we handle people people who don't use IVF because of religious reasons?
Perhaps adoption would be the final fallback, but even then it would still disincentivize people from adopting older children where their previous parents still probably have more of an impact on them than you.
Finally using AI for actually useful problems. Figuring out where the restaurants with the most attractive people are.
The anti-vaxxers in this video are anti-anti-vaxxer. Each person comes up, rambles for a while, everyone else can hear they’re not making sense, and they’re nearly voted off before Dr. Mike can even speak. The anti-vaxxers are intelligent enough to see that the other person’s ideas make no sense but sit down and don’t make sense themselves.
Also disappointing the majority of people couldn’t cite anything to support their points. Did no one even do research to find studies to misrepresent or misunderstand? The majority sound like a regular boring conspiracy theorist rambling about Rockefeller, Bill Gates, and the Jews.
My new favorite pastime is reading the Signal app reviews on the App Store.
“Principled” Anti-Government people when the government starts shipping off foreign looking people with tattoos.
Weird realization. Trump controls the nukes and the entire US army but can’t own a pistol. The fact that every felon is deprived of their second amendment rights seems wacky. Regardless of whether you think Trump should have been convicted it’s hard to imagine that someone who commits business fraud is really a physical danger to society and needs to be deprived of their right to own a firearm.
Take Martha Stuart for example, she’s a felon after convictions of obstruction of justice and perjury in 2004. Again, can you really make a case that she’s so dangerous that she needs to be deprived of her rights?
We already make these kinds of distinctions with misdemeanors. Some misdemeanors are worse than others and should be treated as such. For example the case of Lange v. California said that police can’t always chase a fleeing misdemeanant into their house because some misdemeanors are incredibly minor (like littering). A minor and non-violent offense is not enough reason to override your rights.
Why don’t we make this distinction with felonies? Obviously if you are violent then that’s a great reason to take away your right to own a gun, but felonies are a broad spectrum and it doesn’t seem right to take that standard all the way to the non-violent end of the spectrum.