Hallucinations have been much less of a problem with Astra, so I was surprised by this. I was driving so I uploaded the menu and discussed what I should order. I said I wanted a “memorable drink” from the menu and it said the Taquila, Jalapeno Passion Fruit would be ideal. Only thing is, that’s not on the menu!

ASSOCIATION FOR HUMAN CHESS Statement on OpenAI’s October 6 Release of Chess Analysis Yesterday, on October 6th, 2026, OpenAI, which is currently defending lawsuits involving allegations of plagiarism, copyright infringement, and trademark dilution, released a repository of analyses purporting to contain improvements on a number of high-profile positions in chess. Grandmasters did not ask for this work to be done. The Advisory Group on Chess and Artificial Intelligence, from whom OpenAI has claimed to derive its legitimacy, opened its initial statement by emphasizing that frontier AI corporations should not test advanced chess positions on internal models, particularly without first consulting the players who have spent decades studying those positions. In ignoring the central premise of the Advisory Group’s position, OpenAI has demonstrated total disregard for the norms of chess scholarship, norms established over generations by Steinitz, Nimzowitsch, Botvinnik, and countless other players who understood that the truth of a position cannot simply be calculated. It must be earned. Chess has a particular vision of progress, informed by history, tradition, and field-specific considerations. As Nimzowitsch demonstrated, the threat is stronger than the execution. OpenAI's decision to execute its analysis before the chess community could adequately consider the threat is therefore particularly troubling. We further note that many of these analyses appear to have been conducted without the participation of a single titled player, without the customary exchange of theoretical novelties, and, in several cases, without any apparent consideration of whether the resulting positions would have been aesthetically pleasing to Tal. Kasparov spent years demonstrating the importance of human preparation. Fischer demanded that chess be treated seriously. Capablanca warned of the dangers of excessive theory. And Bronstein taught us that the beauty of a position cannot be reduced to its evaluation. We reject the suggestion that centuries of human insight can be replaced by a number such as +0.73. Particularly concerning are reports that the system has identified a forced draw in an endgame previously believed to offer practical winning chances. Such discoveries threaten to undermine not only established endgame literature, but also the livelihoods of authors who have devoted entire books to explaining why the position is extremely difficult. Releasing more than 700 files at once is not a demonstration of scholarship, but a demonstration of power. The fact that some of these files may contain correct moves does not address the deeper ethical question of whether those moves should have been discovered in this way. As Tartakower observed, the mistakes are all there, waiting to be made. We believe they should continue to be made by humans. We urge grandmasters, international masters, chess writers, and the broader public to approach these claims with due skepticism, to discontinue their work with OpenAI, and to return to a vision of chess that centers human understanding. The burden of proof must remain on those who would disturb the established position. Until such time as an appropriate international advisory committee has reviewed these analyses, we recommend that players refrain from incorporating any purported improvements into their opening repertoires. We recognize that this may result in objectively inferior play. But chess has never been merely about finding the best move. Association for Human Chess Communications Working Group The position is unclear.

I'm pretty sure if I had made people guess before the reveal a lot fewer would have gotten it right. $4,400 vs. $111,500 Morgan Dollars: Can You Tell the Difference? youtu.be/Ngpywpn2JPY?si=TtOL… via @YouTube

The future of coding is more orchestra conductor than hands-on coder. For example: My AI agents keep making the same mistakes over and over. I read this paper: arxiv.org/abs/2608.27454 And decided that instead of self-policing, I’d have another agent write them a wiki based on their learnings. It’s a little more complex than that, but that’s essentially the method. They still make a lot of bad assumptions and mistakes, but this has cut them down tremendously, and they tend to be new ones instead of the same old ones over and over again. I like to say, “Training overcomes prompting,” meaning that over time, the AI will forget what you told it and fall back on its training, even when that training is really incorrect. That’s still true, but this wiki method makes it less likely to fall back on its training instead of following your prompt.

Only statistics minded numismatists will appreciate this one: Do CAC Stickers Really Mean Better Coins? We Tested 122,000 Morgans youtu.be/MqEBAD_znOI?si=f4dG… via @YouTube

Is Honesty a Competitive Disadvantage? lowryonleadership.com/2026/0…

I just published Wishing Vs. Wanting medium.com/p/wishing-vs-want…

I just published When Your Life Rings True medium.com/p/when-your-life-…

I just published Sin Your Own Sins medium.com/p/sin-your-own-si…

I just published Most People Would Rather Stay in Prison Than Try to Fly medium.com/p/most-people-wou…