The details of understanding are a lot less work than recoding, although the coding can be a path to deeper understanding.

Valid blocks are just the longest chain, it is easy to follow the hashes and the difficulties of blocks through and make the same decision about block difficulty. You'll have to use the full and exact rules that bitcoind does when examining every block and transaction though if you aren't using a vetted-by-bitcoin blockchain. If you have a network client that doesn't do any deep inspection, it could easily follow a low-difficulty, invalid rule chain. From the last fork, we see even the smallest details of implementation have to be replicated.

If you want to verify transactions and hold the UXTO set, then you need to build a database, and work through every transaction ever. You'll have to verify the ECDSA signature, and run and validate the scripts. Rules changed at particular blocks, you'll need to reproduce BIP checkpoints and the code that checks them.

You can trust what's in a saved-by-bitcoin blockchain and just parse it. The next step in complexity is called "rewrite Bitcoin".