But in theory type checks could be cached too. Why this isn't currently done is because it's probably more work and requires very deep knowledge of the typescript language and parsing. Meanwhile the bundling itself is a simple affair of combining text files together. So these tools instead just fork the typechecker for some minor gain but in practice you could absolutely cache type information. That's what every IDE does with their intellisense database. I think in this thread we are essentially talking back and forth over the current state of bundlers and most here rationalize why it's okay (if not better!) for type checking to be separated from the compilation but the reality is that it's a fairly arbitrary state of affairs. It could easily all be efficient so that type checking is just as parallelized and incremental as bundling and therefore requires no separation into a different process. As we know, after all a binary produced by c++, rust or c# is also a "bundle".

Actually this painfully reminds me why I found it so weird that tsc doesn't simply offer bundling itself. Why wouldn't it? It should be very easy for the compiler to do this as it has all the information and on top of that, tsc also has an incremental mode already. That definitely means 'incremental' for type information.