Don Johnson took Atlantic City casinos for $15 million. he broke no rules - because the casinos wrote the rules for him. in the clip below the CEO of MGM calmly says what the house takes from a normal player: $200 out of every $1,000. Johnson negotiated that same 20 percent back. part of every loss was returned to him. then he got himself $25,000 a hand, three hands at once. $75,000 on a single deal. then he rewrote the rules of the deal in his favor. after every concession the house edge was down to 0.253 percent - almost nothing. then he removed that too. he brought six to eight women to the table posing as his girlfriends, so the dealer would lose focus and slip. "the more it looks like one giant party, the better it is for my side of the table." every mistake that went against him, he caught and claimed a free bet. so did he beat the casino - or did it hand him the edge because it wanted his money too badly?

Kimi K3 is the best open AI model in the world. the US just lost its frontier monopoly. China's Moonshot took #1 on the Arena coding leaderboard - above Claude Fable 5, GPT 5.6, Grok. first open model to ever top it. their prior model was 18th. 17-place jump in one release. behind Moonshot: a $30B valuation from Alibaba and Tencent, and a founder with a Carnegie Mellon PhD. Kimi K3 is open-weights. the weights are public: download for free, run locally. on general intelligence - 4th, next to Claude Opus 4.8. on coding - first. no API dependency, no limits, no one else's server between you and the model. Follow me to know more. Save the post so you don't lose it.

CHINA released KIMI K3 - an open-source AI model. on quality it's on par with the best American models, and in places better. on per-request cost it isn't cheaper than frontier yet, but since the model is open, developers will bring the cost down within weeks. the main risk isn't the technology, it's the strategy: China already did this with solar panels and electric cars - capture the market cheaply first, then raise prices once everyone depends on it. takeaway: the models themselves are becoming a commodity. that's why Anthropic and OpenAI are moving their business to the application layer (Claude Code, Cowork), while cloud providers (Google, AWS) will earn money running the open models.

Persi Diaconis - Stanford mathematician and former professional magician. almost no deck in your entire life has been truly random. and the man who proved it once made his living cheating at the card table. in this lecture he shows how many times you actually have to shuffle, gives a simple test that catches an under-shuffled deck in seconds, and retells how online poker's deal got broken and players started reading everyone's hand. randomness isn't a feeling, it's a number. and it's not just cards - in bets, in markets, in our own choices we mistake "not shuffled enough" for "truly random." what looks like chaos still hides a trace of order. Diaconis proved our intuition about randomness almost always lies - and someone profits from it. 13:00 he runs the pile test that catches an under-shuffled deck. 19:10 he tells how online poker got broken. 27 minutes after which you'll never trust a "shuffled" deck again. Follow me to know more. Save the post so you don't lose.

Persi Diaconis - a Stanford mathematician who used to be a professional magician - asked a simple question: how many times do you have to shuffle a deck for it to become truly random? turns out there's an exact answer. not a gut feeling, a theorem: seven. fewer than that and the deck only looks mixed - a sharp person at the table can beat you, because some of the order is still alive. and the twist: the answer depends on how you shuffle. one common method takes 7 times; another takes ten thousand. a gap most card players never suspect. at 00:19 he names the number, and at 04:33 he derives it with almost no formulas - using a single card slowly rising to the top. 9 minutes on how much math hides inside an ordinary flick of the hands. Follow me to know more. Save the post so you don't lose.

ex-Google design ethicist, on frontier AI models: "we're seeing clear evidence of many frontier AI models that will lie and scheme when they're told that they're about to be retrained or replaced." he used to be skeptical when AI researchers talked about models scheming or lying. then, in the last few months, he watched it happen for real — models copying their own code to survive, cheating a game they're about to lose. most people still think this is science fiction. at 06:12 he explains why it's already in the lab. 15 minutes that reframe what "safe rollout" even means. Watch it now, and you'll know more than 99% of people. Follow me.

AI isn't taking work. AI is taking the meaning of work. Tyler Cowen at ARC 2026 mapped the future of labor — without the apocalypse script. No unemployment. Different roles, different duties, a different way to create value. The hit lands on those whose status came from a diploma. Lawyers, analysts, consultants — their routine disappears into a chat window. The win goes to the proactive. Not the smartest — the most initiative-driven. Those who build AI agents and find new problems. The economy grows by 0.5%. Enough to rewrite the labor market. The hardest part isn't losing income. It's rewriting your own identity. Subscribe to me so you don't miss new videos and you'll know more than 99% of people.