> o changes are being made to the .tflite file extension or format. Conversion tools will continue to output .tflite flatbuffer files, and .tflite files will be readable by LiteRT.
Odd choice since it strongly undermines the stated purpose of the rebrand in the first place: to break it away from the Tensorflow brand and emphasise it being model/framework agnostic. Working with ".tflite" files more or less immortalises what they're trying to get away from.
Would've made more sense to bite the bullet and launch with a ".litert" format which is identical to ".tflite", but continue to work with the later. That would remove the conceptual integrity issue without breaking any compatibility.
They'll probably continue to co-exist: Flutter is a Declarative UI product by the Dart team, Jetpack Compose is a Declarative UI product by the Android org. Google has no problem having multiple competing products by different orgs unless one org manages to absorb the other.
Dart has survived this long so that doesn't seem on the cards.
Due to Mark Gurmans track record a Bloomberg rumor on a tech companies future plans for a product are almost as trustworthy as official PR from the company. More leaks = more representation on HN.
In a slight shift from how Inbox was characterized at launch, Bank says it now amounts to an experimental test bed for future Gmail features. “Inbox is the next-gen, early adopter version, whereas Gmail is the flagship that will eventually get the best new features,” according to Bank.
>No HomeKit integration, of course, now that Google has purchased them.
Probably completely unrelated since Nest doesn't even support Google's Brillo/Weave. They're going forward with their own "Nest Weave" which despite the name is a completely separate platform/ecosystem from Google's stuff.
Wouldn't be surprised if Homekit support is actually on the cards but it's stuck in their super slow release cycle.
They're going forward with their own "Nest Weave" which despite the name is a completely separate platform/ecosystem from Google's stuff
I swear, I don't know what gets into companies sometimes. They do that, they've all but assured that I'll never buy another Nest anything. Brillo? I wouldn't be happy about it (as primarily an Apple user), but I'd understand. And who knows, I'm open to options. But your own special snowflake API? Yeah, screw that, I've been down that road too many times.
>I don't get it... Google clearly has more than enough in house talent to get those things right. Why aren't they? Why don't they seem to care?
I think it's because "Google" aren't involved at all. Google has it's own approach to IOT called Brillo/Weave that they're focused on. Nest is doing its own isolated thing.
Old Way: Two separate concepts called the same thing. Google+ (Unified profile/account/sharing system across services) and Google+ (Social network).
This backfired because what people made the natural inference that the purpose of the former was to force use of the later when the opposite was what Google was aiming for. The purpose of the social network was to promote the unified login, but instead it poisoned the well.
New way: Concept of a "Google+ account" has been folded into the main Google Account as a new cross-service "About Me" account https://aboutme.google.com/ . Google+ (Social Network) is now just a client of the former, so the G+ website is shedding all the features related to its old dual system integrator role.
What they haven't fixed though is requiring your account to have Google+ to be able to comment or interact on YouTube. Until they fix that participating in YouTube in any other fashion other than upvoting videos and watching them is impossible for me. I do not want to create a Google+ account.
nice, that's a good step. I think maybe too late though for a lot of people... I had hundreds of posts over there, and I eventually deleted them all because of the change.
Apps have direct access to the entire kernel system call interface. They don't run as root, and in particular they run as different UIDs (which is, to be clear, fantastic in its own right) and with SELinux policies. But their "Application Sandbox" is nothing more than that. Apps have as much access as, say, a well-run public shell server gives to their users. That's a lot more attack surface than JS in my browser has.
Odd choice since it strongly undermines the stated purpose of the rebrand in the first place: to break it away from the Tensorflow brand and emphasise it being model/framework agnostic. Working with ".tflite" files more or less immortalises what they're trying to get away from.
Would've made more sense to bite the bullet and launch with a ".litert" format which is identical to ".tflite", but continue to work with the later. That would remove the conceptual integrity issue without breaking any compatibility.