If they planned to use a unified codebase for Prime app, they likely went with something HTML/CSS-based, which would explain the performance issues. I could be wrong, but it's just a hunch I have.
If there was other apps we use that had the same issue, I'd chalk it up to hardware too, but maybe they simply don't test it on representative hardware? That might explain it.
> they likely went with something HTML/CSS-based, which would explain the performance issues
Not sure, the web browser in the TV seems to handle things just fine, and much faster than Amazon's app, so I don't think HTML/CSS is to blame here. Probably shit architecture/software design, as usual.
Could be. Interesting anecdote on that, we're using the Vodafone TV app on the very same TV, and that app you can toggle "send analytics to Vodafone" on or off in the settings, which of course defaults to on.
At one point I toggled it to off, and suddenly the whole app became as fast as all the others, while when the toggle is on, the application is as slow and laggy as the Amazon one. So that might actually very well be the reason.
An engineer from Netflix wrote a blog post in 2017 explaining how they handle LRUD input and focus: https://netflixtechblog.com/pass-the-remote-user-input-on-tv...
They finally agreed to send us the TV. Solved the issue in 10mins.
Normally I'd just use the AppleTV, but the kid stole the AppleTV to watch cartoons in the guest bedroom. I continuously surprised that a 10 year old AppleTV still a better option than using the apps that comes with the TV.
From a developer's perspective, it's a nightmare to deal with such hardware.
Then I got an appletv+ subscription, and was pleasantly surprised it performed far better, on an android tv even. I wonder if it's beyond just the company standards for performance, and that the lower compatibility for porting between swift and the android sdks compared to idk react components or flutter, forced them to start from scratch for performance on android tvs.
> On TV, input works very differently. Users navigate with a remote. Movement is discrete. Every interaction requires intention. Each action is one step in a sequence. That difference changes everything.
I just don't want to read articles that are written by LLMs. If there was something you earnestly learned that you think other engineers could benefit from, use your own words to tell us. It's lazy and disrespectful to hand an audience a massive sloppy blob which reeks of GPT 5 and frame it as something you "learned".