A former physicist turned bioinformatics coder shares how hands‑on work, not coursework, taught him software design. Early formal classes felt like kindergarten, but leading the IntelliJ Rust project forced real architectural decisions. Mistakes on that codebase proved valuable, showing that an inquisitive mind can derive solid practices from first principles and scattered blog posts.

He stresses that code is secondary to the social structures that produce it, invoking Conway’s Law to explain why scientific projects often lag behind industrial ones. Incentive misalignment—such as a PhD deadline—shapes tooling choices more than technical expertise. Adjusting or nudging incentives, though rare, can dramatically improve outcomes, as seen in the development of rust-analyzer.

The author recommends pragmatic resources: Gary Bernhardt’s *Boundaries*, Pieter Hintjens’ writings on organizational impact, and Jamii’s meta‑coding reflections. Classic texts like *Software Engineering at Google* and Ousterhout’s *Philosophy of Software Design* remain useful but not transformative. Ultimately, consistent practice and awareness of team dynamics drive architectural growth.