Rather than write crypto, what I'd actually recommend is to break it. Schneier put together a block cipher cryptanalysis course a long time ago and while I don't usually recommend his crypto books these days, the course is good: https://www.schneier.com/academic/archives/2000/01/self-stud... (in this case, his crypto book might actually be useful, because it documents some of these out of date ciphers. There's (was?) a mistake in the DES code though iirc).

It is essentially a tour of all the early cryptanalysis literature, complete with suggestions of ciphers to target (e.g. FEAL). This will give you a sense of how the attacks work. Many of the ciphers are old, but I wouldn't let that put you off.

The study technique for this would be a) implement each cipher with toggles to control rounds and any other features, then implement attacks. Most of the papers should be open access by now since the 'course' was written in the year 2000. You could also 'catch up' on the constructions and attacks that have come out since.

I would caveat this with: what I am advising is purely for potential interest. Bear in mind there is little need to implement new block ciphers these days (what I'm saying is: this is a very specialized skill and most people won't find jobs in it).