What I don't understand is "dumb" smoke detectors, whether on mains or 9v battery, have been reliable forever. Yet a device that should be not much more than putting a little compute power/wireless connectivity/etc in the same enclosure as a sensor setup with decades of proven reliability somehow fails on the sensor's reliability.

Is there a patent in this space or something that forced them to reinvent the sensor or should this be a poster child for NIH syndrome?

I don't think it's true that regular smoke detectors are inherently reliable. In newer apartment buildings that I've lived in, all of the units are hooked into a central fire alarm system. If a smoke detector goes off in one unit, the alarm goes off in all of them.

In practice it's often been a nightmare. Some places that I've lived, the fire alarm went off multiple times a day, almost every day. Presumably a lot of this is from people cooking or something, but given the frequency I don't think that could explain all of it.

I think in non-networked smoke alarm situations, people are just less aware of the false positives they generate.

And the power supply issue is well solved, too -- see just about any modern phone charger. I just...don't understand.

Conventional smoke detectors use radioactive isotopes (americium). I imagine Nest wanted to avoid putting radioactive materials in their shiny consumer products.

Some do. Many others also use the photoelectric smoke sensor that Protect does, and some use both. The photoelectric type is better at detecting smoldering fires. The ionization (radioactive) type is better for detecting flames. Perhaps with a temperature sensor in the Protect, they thought they were covering their bases.

That's a shame, since the only place a Nest Protect apparently belongs is a nuclear waste storage facility if that's what it takes to get them to shut up after a false alarm.