Planning a travel day in our fifth wheel used to take me about two hours, and none of those two hours were interesting.
It is not the same problem as planning a drive in a car. Our rig is tall and long, so a route a sedan would take is sometimes not a route we can take. We want to stop roughly every hour, because sitting for four hours straight is how a travel day turns miserable. Fuel matters more than it does in a car, and so does whether the place we stop is somewhere we actually want to be for fifteen minutes. On top of that, campgrounds have check-in cutoffs, so the real question is never "how long does this drive take." It is "what time do we have to leave."
Working that backward by hand in Google Maps, dropping waypoints one at a time and rechecking the total after every change, is exactly the kind of task I complained about in how I use Claude Code and Cowork: well defined, fiddly, and repetitive enough that I put it off.
So I stopped doing it.
What I actually hand over
I describe the rig once. Tall, long, towed by a truck, the things that rule certain roads out. Then I give it the two ends of the day and the constraints that matter: where we are, where we are going, that I want the option to stop about every hour, and when we need to be parked.
What comes back is a full Google Maps route with the stops already placed, and the thing I care about most, an answer to the question I was actually asking.
For this Saturday it told me we need to leave by 8:30 in the morning. Not because the drive itself is that long, but because once you add the stops we want to make, the arrival lands against the campground's six o'clock check-in cutoff. That is the calculation I used to do last, by hand, after building the whole route, and it is the one that determines the entire shape of the day.
It also tells me things I would not have bothered to work out. Which of the stops along the way is the better place to refuel. Which ones have food worth stopping for rather than just a parking lot. Those are the details that used to make the difference between a good travel day and a long one, and they were the first thing I dropped when I was two hours into building a route and just wanted it finished.
Why this one stuck
I have been fairly consistent about where I think agents earn their keep, and this fits the pattern exactly. It is not a creative problem. It is a constraint problem with a lot of tedious lookups in it, and the answer is checkable. When it hands me a route I can look at it. If a road is wrong I will know, and I still read the whole thing before we drive it.
The part that surprised me is that the output is not just faster, it is better. My hand-built routes were worse because by the end of the second hour I stopped caring. I would take the first arrangement that worked instead of the best one. An agent does not get bored on the eleventh waypoint, so the version I get now is the version I would have built if I had unlimited patience, which I do not.
That is the honest case for this kind of work. Not that it does something I cannot do, but that it does the whole thing at the same level of care from the first stop to the last.
Her name is Claudia, and it turns out that makes sense
My wife Fabiola started calling it Claudia, on the grounds that Claudia is a better name than Claude. She says it constantly. On our channel she has taken to describing it as my assistant, which is how it ends up in this week's Nomads Amor video, where she tells everyone that Claudia helps me plan.
I assumed it was just a joke about the name. Then I thought about the Portuguese.
In Portuguese, artificial intelligence is inteligência artificial, so the abbreviation comes out reversed from the English. They do not say AI. They say IA. Put that on the end of Claude and you get Claudia. She landed on the correct Portuguese name for the thing by accident, which is a better outcome than the one I got by thinking about it.
What I would tell another developer
Look for the task where you already know the answer you want and the work is entirely in the lookups. Route planning is a good example because I was never uncertain about what a good travel day looked like. I just was not willing to spend two hours a week assembling one.
The rule I keep coming back to holds here too. I define the constraints going in and I check the result coming out. Everything in the middle, the part I used to spend the afternoon on, is the part worth handing over.
If you want to see what the travel days look like from the other end, Fabiola and I put them on the Nomads Amor channel, and the rest of what I build is on my projects page.