When most developers want to scan their code for security vulnerabilities, they install Semgrep or Snyk and call it a day. I did the opposite. I built one from scratch. Not because the existing tools are bad — they're excellent. But because I'm transitioning from 13 years of software engineering into application security, and I wanted to understand what a SAST tool actually is underneath the hood. What decisions go into building one? What tradeoffs do you make? What does "language-agnostic" really mean when you have to implement it yourself? This is the story of those decisions. Some were obvious. Some I got wrong the first time. All of them taught me something.…