Developer Creates Three-Phase AI Coding Workflow That Separates Research, Planning, and Implementation
Summary
Developer Boris Tane revolutionizes AI coding with a three-phase workflow that forces Claude to research and plan extensively before writing any code, using annotation cycles to perfect implementation plans and turning actual coding into mechanical execution of pre-approved strategies.
Key Points
- Developer Boris Tane uses a three-phase workflow with Claude Code that separates research, planning, and implementation, requiring Claude to deeply understand codebases and write detailed research reports before any coding begins
- The core principle involves an annotation cycle where Tane reviews and adds inline notes to Claude's implementation plans in markdown files, iterating 1-6 times until the plan is perfect before allowing any code to be written
- Implementation becomes mechanical execution of the approved plan within single long sessions, with Tane providing terse corrections and maintaining control over architectural decisions while Claude handles the coding work