We start off with the silly circus of the House Judiciary Committee's hearing earlier this week which wasn't worth anyone's time, but we do give equal time to the thoughtful, nuanced letter by academic experts that try to teach Jim Jordan how the DSA is supposed to work...

From there, we look at a new piece in a prestigious education journal in which a bunch of respected academics rip Jonathan Haidt's book to shreds, and point out how it's all moral panic. Ben compares the book to the kinds of content that Haidt claims is damaging to kids...

Then we have a trio of stories on AI. One about an elderly mother in rural China relying on DeepSeek for medical advice (both why it's not great for that, but also why it's understandable it happened). Drs can be curt and condescending, whereas the LLM is neither, and patients want to be heard.

Then another about people making AI "deadbots" to represent the deceased and... how society is not prepared for that at all (though feel free to make a Masbot after I'm gone).

Finally, we talk about how Anthropic and OpenAI agreed to give each other temporary access to each others LLMs for safety/security testing, which is somewhat unheard of and fascinating as an evolution of safety testing. Lots of fun conversation and good to be back at it!