if tesla’s can drive themselves to the person that bought it then presumably they can also drive themselves back if there is a late payment or the car needs to be repo’d. this could theoretically lower the risk for any lending. there is no need for repo men.

i wonder if the deteriorating relationship between openai & apple will cause issues for chatgpt health product given a large amount of health related context lives in apple health app. although chatgpt is a huge surface for health queries too. interesting to see how that will play out.

the future interface is probably three layers: 1. ambient intent capture voice, location, calendar, screen context, messages, habits, biometrics, etc. the system understands what you’re trying to do before you explicitly “open” anything or augments your intent deeply. 2. agentic execution the actual work happens through agents operating software, apis, browsers, documents, email, calendars, workflows, payments, support systems, whatever. most “computer use” becomes machine to machine clerical labor. 3. ephemeral verification ux humans still need to inspect, compare, approve, edit, reject, or enjoy things. that’s where gui survives but as disposable, task specific surfaces generated for the moment.

you can basically tell a lot about a person & their current mood by: - listening to their discover weekly - scrolling through their tiktok feed - looking at their insta explore page potentially even deeper than what their closest friends even know.

i always operate under the assumption that things are easy until proven otherwise. mostly cuz the people loudly insisting things are hard are usually selling difficulty (consultants, gatekeepers, or incumbents who benefit from the moat narrative). also starting from easy means you actually attempt things. the cost of a false easy is often an afternoon now esp with ai whereas the cost of a false hard you never try at all. e.g. this is how i got into woodworking & hard edge painting, among many other things.

the best move is to surface the boyfriend question early but never as a direct probe. fold it into something topical. she says something & you go "haha, i'm curious what your boyfriend makes of that." if there's a boyfriend, you find out in minute 5 instead of hour 2. if there isn't, she'll tell you & now she's the one who volunteered it. you’re welcome.

both chatgpt health & finance products leave so so much to be desired. they’re fundamentally misconstructed imho. forget the model messing up, it’s just kinda frustrating that everything by default is delegated to a chat first interface. chat works well in open ended queries like search & curiosity but absolutely sucks in other categories where the context is broadly familiar but not *clear*. e.g. in health & finance, ppl rarely want to cross examine a blank text box about their own life. what they want is a system to know the baseline, watch the deltas, & flag the weird stuff. most of the time i want to forget shit like this & just live my life. i guess it all makes sense cuz when you’re a bicycle company, every problem looks like a bike problem even when the answer is obviously a car or even a bus. again chat is great for exploration. it’s atrocious as the primary ux for ongoing life management. the real consumer leap in ai software is actually replacing the need to ask in the first place.