Yes and it can also remind you to do something when you "get in the car" or "get out of the car", if you have it paired to your Bluetooth in your car.

But what none of them can do is what people actually need, which is "remind me to do X before I leave work".

What signal could it pick up on before you leave work besides when you get in your car or after you leave somewhere?

Sudden increase of movement (via the accelerometer) around the time they usually leave work, for example.

If I usually leave work around 5-6 and I start packing up my bag, that could be a signal.

Would you use a Markov model for this?

Interesting. I guess I thought that discussing how to model a software problem would be an acceptable and interesting conversation on HN. Do any of the downvoters care to contribute in a productive way? I'd be curious to hear what people found disagreeable in my question.

It could guess and or extrapolate, based on recent data.

For a lot of people it would be effective.

Nest is already pretty effective at learning schedules

Nest doesn't have to be accurate to the minute to be effective. A reminder based on before I leave work would be.

For instance, if I don't know what time aim getting off of work and I wanted a reminder to call my wife when I got to the car, it could do that completely accurately based on phone connecting to the car. It could only guess about doing something before I left work.

Yes. "Hey Siri, remind me to bring a change of clothes when I leave home." just now created an appropriate item in Reminders.

which will, of course, trigger the reminder as you walk down the street away from home, forcing you to go back and get the change of clothes that you forgot.