If there's a fire, your insurance company will likely be able to escape issuing a payout if they discover you knowingly brought about a fire code violation.
I would actually do this, but use Home Depot smoke alarms that are up to code as well. The custom smoke alarm would only be for the sake of alerting me remotely or something.
Imminent danger of death? I want a Nest (that works, though this video is anecdotal and seems staged-ish) as much as the next person, but I have to silence a standard smoke detector once every never. It's inconvenient when I do, but less inconvenient than waking up to find my house is on fire or not waking up at all because I died in my sleep from carbon monoxide or smoke inhalation. Even if I trust my own build and code, do I trust the build/code quality of an arduino and whatever half assed, 3rd party smoke detecting shield I can find over a device designed and built by people with the life saving responsibility in mind? I'd still trust a Nest over any arduino shield.
What if the home contained both normal per-regulations smoke detectors that alerted humans, and had other smoke detectors to alert your home automation system? We could call them sensors instead, to differentiate.
Ideally, the sensors for your home automation system would be configurable to your hearts content, their point is not to alert humans but to alert your home automation systems. The actual smoke detectors are for alerting humans, and are tightly regulated.
I see a future where the home contains many different types of sensors all over the house intended for data collection, and alarms for humans as well. Temperature sensors and heat sensitive cameras combined with smoke detection sensors would let you do some really interesting things with regards to detecting smoke, flame, or even overloaded extension cables.
One of the reasons I'm interested in my home itself being able to detect events such as fire or flood, is so that it could initiate an emergency offsite backup of any sensitive recently-modified files, and as soon as that completed or failed for whatever reason it can shred the encryption key used to boot the full disk encrypted partitions. It could also unlock doors, trigger a gas and or electrical mains cut-off switch, trigger camera feeds to record to offsite servers, call me or anyone else, or even draw attention to the home by turning on and off lights (or changing their color if you have the right bulbs).
Personal risk aside, have fun trying to get your insurance money if something goes wrong, even if it's not the fault of your home-built fire detectors or Raspberry Pi thermostat or whatever. I'd be surprised if that didn't cause problems, should you have to make a claim they find out about your tinkering.
The notion that this is an either-or situation, where a device must necessarily be prone to either false alarms or failure to alert during a true emergency, makes me hope that you (and most of the other posters in this thread) don't actually work for companies that make these things.
I don't... and have tremendous respect for people who build software and systems that are "mission critical."
They're in the same situation as civil engineers and the like. They build a bridge that falls down, they go to jail or are financially ruined. We make software that falls over and people pay us for updates.
