Let me ask this question. Why can Netflix make a decent tv app and Amazon cannot?
[0] - https://www.hollywoodreporter.com/business/business-news/apo...
[1] - https://variety.com/2025/tv/news/prime-video-india-growth-pa...
[2] - https://variety.com/2025/tv/news/prime-video-india-content-c...
(I know the literal answer is on Wikipedia, but I’m flabbergasted.)
But they don't really think about things from a "consumer who wants to watch something tonight" vs "shopper who we want to get money from" perspective. So the Prime Video app has been painful to navigate and use. Things like concepts of how people want to interact with TV shows - one top level entry with seasons in it, vs top-level entries per season, which took them forever to change - reflect quick and dirty shoveling of concepts over from how they'd sell box sets or such vs thinking about it from a user-first POV. Or how search will return a match for just about anything because they will happily sell it to you vs having as a default "show any free results first because I'm not looking to spend more right now."
That's a product/vision failure (or just mismatch with what you and I want) not an engineering/engineering culture thing.
- Amazon has 3 main business lines ("orgs"): Ecommerce, AWS and devices.
- Ecommerce and AWS are (now) cash cows. Devices bleeds money. TV falls into the devices organization.
- Devices was a Bezos bet. Current Amazon couldn't care less honestly.
- The devices organization is (today, after layoffs and people leaving in droves) essentially full of incompetent people, where all the leftovers of the other two orgs end up.
- It's people that was hired to build structure with the sole purpose of some higher-up promotion. They never served any other purpose, neither they have any particularly sophisticated skill.
- That's the people that makes the TV app.
1. scaling retail
2. keeping the servers running at AWS
all the low hanging fruit has been picked, they need a fundamentally different employee base
As anyone in software development can tell you, this does not compute. You cannot do things this way, and any experienced software engineer can tell you it doesn't work.
Besides, it's not how Amazon worked at all. Amazon is famous for having systematically verified ("mathematically proved") how it's core systems operated. Whereas, for example Google only did that in redesigns when the systems had already collapsed once or twice due to scale, not from early on. And even that is superior to how Microsoft or Oracle did it: they bought Google employees and had them design an iteration of what Google is running (yes, is running, not was running. Google redesigned it's core systems ... and then mostly didn't migrate. Borg was never replaced with Omega and the main large system that they migrated to is Spanner. Kubernetes isn't Borg. Kubernetes grew out of Borg's successor. Except Google never migrated away from Borg)
https://cacm.acm.org/practice/systems-correctness-practices-...
I'm sure Amazon had entire departments, much larger than core engineering, just like every other company, where it looked like everything was operationally focused. That doesn't mean core engineering doesn't exist, or does nothing.
The Amazon way is to quickly grind something out, and build on as many layers of AWS abstractions as possible. They’re famous for the hustle and grind not the stellar engineering acumen of formal verification.