Mandatory features of any smartwatch that costs more than $100:
* Acceptable aesthetics.I'm setting the bar pretty low here, but it has to at least look rugged & utilitarian, if not actually attractive. If it looks like a Fisher Price toy or some cheap piece of plastic junk, it's not happening.
* Ability to use normal wrist straps, absent some compelling and good reason to the contrary.
* Glass that's either independent of the screen & can be replaced when cracked by me for $10 or so worth of parts and an hour of time, or hardened enough to survive getting repeatedly scraped against rough concrete walls. I destroyed dozens of watches growing up by accidentally getting too close to a wall/concrete pillar/whatever and scraping or smashing the glass.
* MINIMUM 36-hour battery life
* At least two tactile hard buttons that can be easily pinched independently of one another and used as a modifier key with the other. I hate HATE ***HATE*** touchscreens in general, and a watch would be the worst touchscreen environment of all. The only way to make it random-touch-resistant would be to add latency and sample delays that would make it feel laggy & slow.
* Rootable & reflashable as I see fit. Android would be nice, some Linux variant would be OK, and frankly I could live with an Atmel AVR as long as I can personally reflash it.
* Real, honest-to-god e-ink (not LCD-based "e-paper") display that takes a cue from the DSTN LCD displays of yore & has two or more independent controllers that can update different parts in parallel (doubling or quadrupling the time to redraw the display). Enough framebuffer ram to do full-blown double/triple-buffering with *really fast* DMA (to let you compose changes, then propagate them to the actual display in an instant instead of 200-400ms) would be even better. There's no technical reason why an e-ink display HAS to be glacially slow... they've just been slow up to now because they were designed to minimize component cost and conserve battery life. But since they'd only consume power while being actively updated, the power budget difference between e-ink with parallel controllers and e-ink with one slow controller would be fairly small (think: race to sleep instead of always running slowly).
* If it DOES have a touchscreen (in addition to the aforementioned pair of diagonally-opposed hardkeys from a few points back), that touchscreen needs to be capable of AT LEAST 120 samples/second (if not with stock firmware, at least the hardware itself when reflashed to a custom ROM). A tiny screen NEEDS a high sample rate to get any kind of acceptable resolution from a capacitive sensor.