I'm not sure this is something I want to pursue right now, but I stumbled upon a Reddit comment from [Mike Szczys] some time ago that stuck with me. He brings up a great point regarding utilitarian IoT devices; as an early purchaser and rapid abandon-er of a Fitbit, I'm definitely interested in cheap, connected sensors that have real value.

He used the example of the mandated CO/smoke detector--why didn't it send notifications to one's smartphone?

I started thinking about how I could get around this. Build a smoke/CO detector? Probably doable, provided I could find the parts. It would probably be easiest to use a reference circuit from the various sensor datasheets, assuming that sort of thing exists--maybe the First Alert company (and others) actually manufacture and own the supply chain for their sensors. Who knows?

I think a better approach could be to bypass the alarm guys entirely. Not that there isn't room for a product like that; last time I bought a combo detector I think it was in the neighborhood of $50, give or take. Seems like these units often come with other fancy features (like the dude that yells, "Fire, Fire, Fire" whenever I cook steaks), so maybe there is enough room in the budget to throw in an ESP8266 or something. But then you'd need to worry about the safety side of things--it's a smoke/CO alarm, and your life really does depend on it working reliably. At the very least, you'd want to install an unmodified factory original unless you somehow got the project through certification.

What about an ESP8266, a suitable microcontroller (or itself--from what I understand, folks have figured out how to program them), a microphone, and maybe a dedicated A/D chip? Or an analog bandpass filter, tuned to the audio frequency of a typical fire alarm? Put the system in some type of 'training' mode when it's near the alarm and trigger a smoke alarm test. Could a Trinket Pro handle the A/D part natively to minimize size and board count?

I don't want to spread myself too thin. I just picked a project up that I'd set aside for another project, which I've now set aside. But the Trinket Pro contest could be fun.