Parallelism comes in the collisions. Can you monitor that Hue hub over local IP, do your system maintenance, handle a scene edit while triggering an event based on internet weather reports,run that data logging app the user installed and update the two wall tablets?
If you get the prioritizing right, probably close enough to keep the user happy. But getting that prioritization right is a big if. Can you tell a non-time sensitive data logging app from a plugin that controls a web-based light? How far behind can your tablet refresh lag before it irritates users? Can you delay the system maintenance task or will it cause other issues later?
As for hardware vs software, your web server may use that little ram if you aren’t caching any icons, which pushes the load to your storage. Does more ram make lazy coders or does the search for a mythical efficiency come at the expense of flash life? Tech weenies can argue about ram footprint vs flash life for days, weeks if you add in engineering expenses for optimizing code.
Users will want to know “will this slow down when I add the Kasa plugin?” “If I access a camera remotely, will the lights still respond quickly to the voice assistant?” “If I add this data logger, does the device thrash when six automations trigger at the same time?” “When I have 40 devices does the app make that comprehensible or does the UI go all squirrely from the javascript loop?”
“How long will it take to get support for (new hot thing)?”
I am the skeptical HA curmudgeon who’s going to look at competitors and say “There are four engineering teams who have failed to produce infinite punch and pie, what trade off are you making?”
ou, as CEO/owner, are the corporate cheerleader so you’ll say “Relax, its gonna be awesome! Punch and pie for all!”"
I expect some punch and some pie but not for all. Or, if for all, then for a price that doesn’t make sense. You expect to prove me wrong.
When your products come out we will see if you did give punch and pie for all at a price worth buying.
Until then, I will look at the data in official posts and compare it to your competitors because I don’t have a product to test.
(Semi-off topic: I worked with a programmer who wrote assembly modules for a C based real-time telecom billing system. It was hyper fast…until the SGI hardware it ran on was EOL and it had to be moved to x86.
My point is optimization targets specific use cases. These days the $ cost of more powerful hardware vs opportunity cost of not supporting a wider range of use cases is not an automatic pass.)