New model drops are making me revive old projects. Fable 5.1 is amazing but Astra... A whole new OAI family.
I have frontier math problems waiting.
I've spoken with several educators lately and their views on pedagogy and AI fail to understand how much their field changed.
I think that homework is dead as a technique for structured education. Learning now happens I der supervision. That's ok. Adjust!
My daughter will still use tons of AI, but she will write. All by herself. In a room without a computer.
Sean Cook retweeted
What do you think happened in 2010 that started causing more liberal teens to say they don’t want to have children in the future? Look at that drop! Wow.
Basically every professor or academic has turned off replies. What's it mean when the intellectual class has decided against engagement with the people?
If a science-fiction writer had managed to write an alien species who talked like a Fable thinking trace, it would have been the best example of well-written "alien thinking" in the genre.
I love this clean, beautiful visualization of the election results for San Francisco.
electionmapsf.com/2026-06-02
I gave Opus $40k to trade on the stock market, focusing on the pickjs and shovels of the AI revolution, at aishovels.fyi. Let's see if Opus can crush SPY.
If you're interested in buying auctioned public land, I made a much more browsable version for the site most counties use.
Go buy some 1 dollar land in the middle of nowhere!
bid4asseets-reader.com
My 18 m/o daughter wants to get up despite me having another 30 minutes for coffee and Slay the Spire II per the Deal.
nitter.cf/ColbyBadhwar/status/20…
This is such an interesting video. The Ukrainians had to buy a bunch of random drones, almost all running on Android, and then they jailbroke them and started writing their own software.
The @SecArmy is basically saying "Holy shit, they networked their entire front line and started building their own targeting software."
Now we will tell US Army contractors everything has to be open to, I guess, Army development?
It's very easy to see how the Ukrainian Army got there....it couldn't have been otherwise. I hope we can find the flexibility and capacity to emulate that in a system that operates in a wildly different manner.
🇺🇦🇺🇸 @SecArmy: "What we've learned from the Ukrainians [is when you jailbreak your equipment] you can start to layer generative AI models [on top] to help with decision making."
@SenatorWicker: "How long have the Ukrainians been using this concept?"
Sec Army: "The entire war."
Russia is in an incredibly unstable position.
They're losing 30k men/month and have no functional economy so long as the war continues, but Putin can't stop the war without losing control of the country.
Ukraine can turtle mode for several more years via drone autonomy, while making increasingly visible strikes on Russia.
In Russia, Putin is breaking up internal comms networks to prevent coordination among revolutionary parties, and kill elites to prevent their coordination of a coup.
Elites knows Putin has to go. They are waiting to see who moves first. None of them want to. But the war continues until Putin goes. And the war consumes their country.
Russia is in an incredibly unstable position.
So Russia worked very hard to pollute the American political space favoring extremists rhetoric, Trump gets into power, is nice to Russia...then systematically destroys all Russia's friends.
Putin has a monkey's paw with 20 fingers and they won't stop curling. Guy needs to stop making wishes.
I built a latency monitor to test the "arb Binance vs Polymarket BTC markets" strategy people are selling bots for.
Results: Polymarket reprices in 15 minute markets 0.1-0.8s. Market makers are already there. The book also moves 60+ cents on flat BTC due to binary option gamma.
Your edge per trade is ~1 cent, the volume is tiny and the strategy is dead.
If someone's selling you the bot instead of running it, that's your answer.
Been concerned that people are forgetting how to code is pointless. It's the modern equivalent of being worried about forgetting how to weave tapestry by hand.
Here is a fun comparison of the different safety philosophies of OpenAI, Anthropic, Gemini and Grok when identifying people. And a work around showing how ridiculous this policy is on famous people.
The new players are nervous in the face of uncertainty, Google has aggressively recalibrated their tolerance for risk since launch, and Elon simply does not care.
I think there will be worse long-term consequences to lobotomizing our models than there are to identifying celebrities.
(I would try this with Llama, but I don't want to spend the next 3 days getting ads for Def Jam Records.)