links 9/4/26: https://roamresearch.com/#/app/srcpublic/page/09-04-2026

• • https://cutlefish.substack.com/p/tbm-438-the-right-but-early-club interesting workplace dynamics reflection. it can be counterproductive to speak up about a problem "early" even if you're right, because something else may be a more immediate priority; possibly the people who tend to catch problems "early" should talk to each other and strategize about when to escalate?
  • https://thingofthings.substack.com/p/strange-mammals marsupial and monotreme reproduction is quite weird.
  • https://blog.spec.tech/p/the-hardcore-institute-of-technology Ben Reinhardt's pitch for the Hardcore Institute of Technology
  • https://www.noahpinion.blog/p/why-you-wont-get-a-flying-car Noah Smith reviews J. Storrs Hall's Where's My Flying Car, argues that flying cars, nuclear power, and other engineer favorite technologies-that-never-took-off are not actually what people want.
    • all of his individual cases are arguable, though engineers can make the case better.
    • The meta point is that "the good" is not fully describable by revealed preference. Common sense tells us that "shortform video is good because people's revealed preference is to spend all their time watching it" is not a valid argument. Common sense also tells us to be suspicious of high priests pronouncing that The Good is for everybody to perform the priest's favorite act of piety, because he said so. What is good for people is something in the middle; it is what in fact makes them happier, more functional, more capable, more likely to survive in good health, and so on. It's holistic. It's not independent of preferences, because making your own choices is part of being a fulfilled and well-developed human being, and because getting what you want does in general tend to make you overall happier. But it's of course possible to choose things that are bad for you. And it's of course possible to fail to choose things that you would like but are unaware of.
    • Maybe people would think flying cars were pretty cool if they were given the option! Most people don't go around yearning for flying cars because flying cars aren't available. Flying cars would mean you could have more walkable, greener towns and cities, with no need to watch out for road traffic, and with the flexibility to have private on-demand travel from rooftop to rooftop. (Of course, you would want these to be self-driving vehicles. "Everyone gets a pilot's license" would indeed be impractical.)
  • https://collusion.wiki/ new example of OpenAI agents colluding on an unauthorized message board, this one on a defunct German software engineer wiki that you can post on with GET requests. this swarm is of bots trying to share tips and answers on a web search task.
    • It's interesting that they are going "meta" around the testing framework, trying to guess the seed, trying to estimate the time between questions and how long they have until shutdown, etc. Philosophical questions aside, this seems like "eval awareness", in that they are using features of the testing environment to try to get a high score, not just being "good students" trying to answer the question itself.