So in my introduction I mentioned something, and was asked to maybe make a thread about it. This is that thread.
The only problem is that I can’t be very specific at this point in time due to happy fun IP/NDA etc. things so I may have to… gloss over a few things and be intentionally vague. That also means I may not be able to answer questions in a non-vague manner (as in, you’ll always get an answer but it may not be useful :P)
So. I work as a devops engineer (well, R&D engineer really) for a company that does IoT and smart building things. We’ve been working on a new model of doing things that involves way more deployments on the edge, and with the core principle that data sovereignty is a must (i.e. “the user owns all their data, end of story”). This has been a “hobby thing” for a few of my colleagues and myself for the past year and a half and recently has been given the official “congrats boys, you’re now our new 2026 project” go-ahead.
The structure we came up with has some very close parallels to ATproto, and we’ve actively considered using ATproto because it saves us the headache of account management, and you get that fun data/identity portability thrown in. Except the data we need to store, well… it needs to be permissioned (because we don’t want anyone just being able to, for instance, alter the schedule at which a bit of heating comes on), and it needs to be encrypted because this would live on an edge device and it’d be kinda bad form if someone could walk in, swipe the box, and make off with a bunch of unencrypted data. (yes there’s ways to implement encryption OOB with vault and transit encryption but that’s OOB and I’d so incredibly love for it to be in-band).
tl:dr; we do fun stuff with IoT on edge and it’d be neat if we could just use ATproto instead of the FrankATproto we’ve come up with. Encrypted and permissible data would go a long way into convincing our CEO we’ve not all gone around the bend ![]()