I get their reasoning, but it still saddens me to see them surrender like that.

Pretty much every site that lives off of ads nowadays is "disgustingly anti-user" the same degree that Facebook is. The main difference being that Facebook had the capacity and confidence to wage a full-on war on ad blocking like this. Now, with this method proven by Facebook, it's likely that more websites will follow suit.

It's definitely cool that you can get any reasoning whatsoever out of such a small model. That said, its reasoning is "interesting":

Query: "Make the living room dark" Agent: "User wants lights on in living room. 'dark' implies dim. Room 'living room', action 'on'." (And on every test I did, it just completely ignored the "brightness" parameter)

It also appears to have no concept of what a door or light actually is, whenever the query diverges from "Lock door X" or "Turn on light X", it tries to shoehorn whatever additional context is given into the device name:

Query: "Lock out the vacuum salesman at the front door" Agent tries to lock "front door vacuum salesman"

"The way you talk really makes me appreciate silence" is classified as "positive" with 82% confidence.

Ok, this is genuinely funny, we will fix these as we iterate, thanks lol.

that model is 14MB large what do you expect. but I agree it's funny regardless

What it can actually compile is a subset of well-behaved typescript. For anything else it bails out and asks you to ship a JS engine (albeit, not NodeJS) again.