Heavy AP user here. Both in stop-n-go traffic and road trips. Recently took another 5 hour road trip and except for 2 construction zones and a few blocks of city driving (to Supercharger + food), AP "drove" the whole time. Probably 4:50 of the 5:00 hours. Arrived refreshed feeling like I had only driven an hour.
AP did brilliant. Never straying from it's lane? Check. Virtually no ping-ponging? Check. Handled cresting of a hill without disengaging? Check. Mountain curves at 75 or 80 mph? Butt-clenching...but still check!
Zero unintended disengagements from AP on that trip.
Obviously I had to stay fully engaged, especially so around high speed curves. But even though you are mentally engaged (and, yes, holding the wheel), not having to operate pedals and steering for 400 miles makes the trip fly by.
I don't love the new/more frequent nag schedule...but it should be obvious to us all that Tesla is doing what it needs to do to reduce the number of AP accidents that get reported by the media that equates Tesla and AutoPilot in a headline with an order of magnitude more clicks than "Toyota Camry rear ends fire truck while on cruise control".
All of this said, it is progressing much slower than Tesla thought/said and much slower than we all want. My local neighborhood has a ¼ miles stretch I use as a gauge for how AP is progressing. In some ways it handles it much better than it did a year ago when I bought the car. But even the very latest (currently 2018.21.9) version cannot handle the first 'S' turn on that stretch without going across the dividing line into the oncoming lane
. To give it "partial credit" I guess it doesn't go quite as far into the opposite lane as it used to at that point. But if there was a car in the oncoming lane, it would still hit it. After it gets passed the first S turn, it handles the rest of the stretch well. Mostly. Except at the very end, which it handles inconsistently. 75% of the time, it handles the last (sharp) turn well and the other 25% of the time, it seems to not 'see' the berm (or the turn!) and wants to barrel full bore into the berm! Obviously I would like it to be 100% consistent but it is not yet there.
And at this point, none of the "Enhanced" features have really/fully materialized (auto handling of highway transitions, on-ramp to off-ramp handling, autonomous lane-change, smart summon). Although possibly by this summer this could change.
Personally I really like AP and while I too am frustrated with the relatively slow progress on it (I say relative because Tesla seems to be working at a faster pace than MobileEye did and also faster than most others players), I do believe it will get there. If you want to spend an extra $1K and just wait a year or two, that's certainly an option if you don't want to be part of the beta process.
Supposedly Tesla will be re-doing the AutoPilot Trial so if you do decide to forgo it, you can probably try it out. But if you admit to yourself that you will someday buy it (even if that is 1-2 years from now when it is more mature), then I guess it comes down to opportunity cost. What will you do with that 5K. If you have a great investment to have that $5K earn more than the $1K post-purchase premium, then that could be an option. But if you are going to take that $5K saved and keep it in a bank account earning 0.25% interest or something, I'd just buy AP now and simply don't use it until you hear it is mature enough. But warning...that will be hard. If you have it, you will want to use it, despite it's flaws and nags.