You don't need to ask anyone else's permission. The thing that got me over the hump was wanting secure chat with an AI agent running on my own hardware. Asking WhatsApp for an API key just rubbed me up the wrong way.

Snikket claims to solve this but assumes it's got a routable IP address. If you've got it on a VM somewhere cloudy you're probably fine but I had to really jump through hoops to get it working behind NAT on my own hardware.

Has someone got a quick explainer for why adding a planet would result in less lithium? Is it just that a super-earth would be big enough and have so little lithium that it would measurably water down the other elements?

Neither an astrophysics. I skimmed the paper. IIUC the collision with the planet mixed internal layers of the Sun and a lot of the Lithium went to the core and got fussed with Hydrogen.

PS: The paper says that that actual A(Li)=1.1 instead of the expected A(Li)=3.3. I needed a lot of search to understand it until Gemini kindly liked this page. https://astronomy.stackexchange.com/questions/61814/what-do-...

A(Li)=1.1 means the Li/H ratio is 10^(-10.9) = 10^(-8.9)%

A(Li)=3.3 means the Li/H ratio is 10^(-9.7) = 10^(-7.7)%

Not an astrophysicist, but since they mentioned the lithium is depleted, I wonder if that implies the collision of a planetary body with a sun would consume or react with the solar lithium.

Billipodes. Obviously.

That sounds like a reasonable estimate for the gap between any pair of Chrome, Firefox, and Safari, just from varied depth of implementation of more recent standards.

This is (mostly) the way. Samsung TV without networking configured, XBox for everything - which is problematic in its own ways but it's a known quantity. It doesn't prevent it from "helpfully" hopping on a nearby unsecured network but a) I haven't spotted one of those in a while, and b) it hasn't ever been able to get updates to change its behaviour to make it start doing that if it previously wouldn't.

Yes, but as a consumer you want to pretend they aren't. Seriously, buy from anyone else.

In my experience the thing that Fable is superb at - unmatched by any other model so far - is downgrading to something else at the slightest opportunity.

> The efficient frontier of LLM inference is a line, not a frontier.

The efficient frontier of portfolio theory is also a line. Not sure I get your point here.

not sure i get why you don't get it. why have the word frontier if it doesn't mean something different than other words?

Because... it's an accurate description of the concept it's being applied to?

It's also the hardest point at which to try to work, because when you change the model architecture you need to completely retrain from scratch.