I know plenty of sharp developers who put off side projects for months because they want the whole architecture clean in their head before opening an editor.
In practice, you rarely figure out a hard problem in advance. You pick a tiny first target, discover three wrong assumptions before dinner, stare at a compiler error you barely know how to search for, and fix one concrete thing at a time.
Don’t start with “I’m going to build a full operating system.” Start with “I’m going to get my own C code to boot and print five letters.”
Almost every project on this site started that small:
- SYPAS started as getting firmware to jump to my kernel entry point and draw text on a blank screen (more on that here).
- Gable started by tokenizing
1 + 2long before structs, lists, or multi-file imports existed. - Tiny Git started by hashing a single file into
.tinygit/objectsbefore commits or merges existed.
Reading ten more articles before writing line one feels productive, yet teaches very little compared to watching a rough v0.1 fail and digging one layer deeper to see why.