Back in the early 90s, I was doing mac68k programming. Very large Inside Mac volumes were pretty much the only documentation for anything - no stack overflow, no internet, no searchable docs. Almost everything was closed source, and many people wrote their own data structures. When programming Macs, if there was any kind of null value dereference or pointer problem the computer usually crashed and you had to reboot.
The main thing I learned was to be able to look at code and determine if it was going to actually do what I wanted it to do without having to run it through the debugger a couple times.
It's important to spend time thinking about and reviewing code when edit, compile, and test cycles start to get long. There are a lot of instances where that still applies, such as with multi-threaded programming, integration testing or long running jobs.
I also learned the value of having muscle memory for APIs. Sure you can look things up in a jiffy, but actually rote memorizing stuff that is used often can speed development up.
The main thing I learned was to be able to look at code and determine if it was going to actually do what I wanted it to do without having to run it through the debugger a couple times.
It's important to spend time thinking about and reviewing code when edit, compile, and test cycles start to get long. There are a lot of instances where that still applies, such as with multi-threaded programming, integration testing or long running jobs.
I also learned the value of having muscle memory for APIs. Sure you can look things up in a jiffy, but actually rote memorizing stuff that is used often can speed development up.