Comment by mitxela
21 hours ago
You're right, "tell me which of these four preconceived routes is fastest" is much nicher than "telly me the fastest route"
21 hours ago
You're right, "tell me which of these four preconceived routes is fastest" is much nicher than "telly me the fastest route"
It’s not the fastest route, because the algorithm is stupid as a result of trying to be smart.
It sees a small side road and thinks “I will take the user down here because there’s no traffic” But it doesn’t consider that the side road is empty because you’ll have to turn across four lanes of traffic in rush hour. So you get stuck for twenty minutes.
How would an LLM know not to take that road then?
If your answer is you tell it, then the obvious solution to improve the app is to provide a way of specifying a road exclusion list, and that is extremely feasible.
So it's wrong about the fastest route. Fix that instead of inventing a new type of app.
You get that this is a commute, right? Not charting a course through unfamiliar territory. A drive I've done thousands of times. I don't care if Waze thinks I should be on Harrison or Race. Fuck you, Waze. There's 4 reasonable routes; I just want to know which one.
It's an app with an audience of just me. That's the point. You're not supposed to want it. You have your own problems.
FWIW I do kind of see your point. I have lots of software / hardware with a user base of one to a handful of people. I also know it sucks having to maintain software that drives your life, even if that maintenance is AI supported.
Just with the bit I quoted above, that seemed like a big misread of the market and the implications of everyone writing their own apps.
In this future, are people paying the API pricing of traffic providers? In this future, who's supplying the traffic data if no one is using their mapping apps?
Apple literally has automatic commute notifications that tell you the optimal route (15 minutes via Easy St.), in this hypothetical world where your agent knows enough about you to suggest today's route, why doesn't your digital panopticon purveyor (be it Apple, Google, your car manufacturer) automatically pull that info the second you get into your car?
I don't know how many different ways there are to say that I do not want "the optimal route" according to Apple or Google.
I think this is coming across as glib or crabby but there's something more significant I'm trying to communicate. This subthread is in one of the classic software developer problem solving modes: pose a question or a problem, respond with "what are you really trying to do", argue that the problem was misconceived to begin with.
Eric Raymond wrote a whole giant piece about this.
But the world in which nerds could congratulate themselves about knowing what the real questions were and separating them from the dumb questions being asked by noobs who aren't holding things right is gone. Why do I care whether a software developer thinks I'm holding the tool right? Why do I care that someone familiar with Apple's fixed-function prefab software ecosystem believes there's a better way to handle my commute than what I'm trying to do?
OK? You use those tools. I'm just going to have my computer do the thing the way I want it to work. It's my commute, and my computer, this is the very specific tool I want, and all it took to get it on my screen and working reliably was for me to ask an agent, which didn't lecture me about the proper way to ask the question or what question I should have asked instead, but instead just fucking did the thing I asked for.
It's wild to me that software nerds believe they're going to win this argument.
3 replies →
we could spend time answering that question or, as tptacek proposes, diy the solution because the app making app is right there, and answers to your input not Apple/Google's