TL-B seems very low-level. Do I need to write a TL-B my own if I'm trying to develop smart contract? In what situation do developers need to understand and write their own TL-B scheme?

Normally, we discuss TL-B associated the storage of data in the TVM, or the way we modify data in the smart contract. If you follow the [TEP74-Jetton] and Tact implementations here, you can easily identify where the TL-B points out in each line.

In my opinion, it is more like an API standard that people define if they need it, so that contracts can interact with each other.

Understand - on interaction with other contracts. By the way, schemes used are often informal - not parsable by programs.

Thanks for ur information. So can I said that the utility of TL-B is somehow similar to ABI but for interactions between contracts?

That's an issue I've been confused for a while. Does the TL-B (.tlb) file actually take part in the process of smart contract deployment? Or just as you've mentioned, not parsable by programs

No, it's rather documentation.

Got it, thanks