What does this mean then

It just means hive specializes in certain types of data for specific applications, and providing convenient APIs that wouldn't make sense as part of the steemd consensus node.

I want to know what the entire system will look like once all these layers are implemented?
If it's all fragmented how can it be considered secure and blockchain based if more and more of the data is going to exist on regular systems?

There's really not all that many layers here, and it makes for a more agile environment. If there was one daemon that did everything, nobody would be able to run it. So now we can have more lightweight engines talking to each other. You could run steemd+hive yourself and it will be less painful than a full steemd node while offering more power. Or you could run hive and sync off a public steemd node.

Because more users can run more lightweight services, I see this as increasing security and anti-fragility. The blockchain remains the source of truth, I don't see the issue with storing copies of blockchain data on "regular systems" for easy access.