Every to-do app comes with a second job
A list only works while someone keeps it. The software worth building is the kind that keeps itself.
Most people I know have started a to-do app at least four times. The first week is lovely. Everything goes in, the list is clean, the day feels held. Then a busy Tuesday arrives, the app stays closed, and by Friday it is a record of things that were true a week ago.
Why do people keep abandoning their to-do lists?
Because the list asks for upkeep, and upkeep is the first thing a busy person drops. Someone has to write the promise down when it is made, open the app to look at it, tick it when it is done, move it when the date slips, and clear it out when it stops mattering. Five small chores, every day, forever.
On a quiet week that is fine. On the weeks that matter, nobody has time for it, and the list drifts away from the life it was meant to describe. Once it is wrong, it is worse than no list at all, because now it is a reminder of everything you have let slide.
People read this as a failure of discipline, and switch apps, and start again. The new app is more beautiful and asks for exactly the same work.
What would a list that keeps itself feel like?
You would say what you need to do wherever you are, in a sentence, and forget about it. In the morning you would see what fits today and nothing else. Things you finished would be noticed. Things you let go of would quietly leave, instead of sitting in red for a month.
The test is simple. If you ignore it for three days, is it still right when you come back? Almost every productivity tool fails that test, because it was designed for the person you are on a good day.
Where does AI belong in this?
In the upkeep, and nowhere near the decisions. A model is good at hearing a rambling sentence and finding the three tasks inside it. It is good at noticing that the call happened and the task can close. It should never decide what your day is for.
That line matters more than any feature. The moment software starts moving your life around without asking, you stop trusting it, and an untrusted list is abandoned faster than a stale one.
I am building one of these now, for myself first. The brief I keep coming back to is the one any good product starts from: find the chore the person resents, and make it disappear.