Lab experiment
Agentic Development
Practical ways to make coding agents useful inside real engineering constraints.
This diagram is a working model, not a deployed architecture.
01
observations
Established practice, open direction
Claude Code, OpenAI Codex, and Cursor are already part of day-to-day investigation, planning, implementation, debugging, testing, review, and documentation. Here I am working through how to make that practice more reliable and repeatable.
02
workflow
Working loop
- 01Context
Supply product, architecture, code, and constraints.
- 02Decompose
Turn a goal into bounded engineering work.
- 03Execute
Use agents where they improve the workflow.
- 04Verify
Check decisions and output against evidence and tests.
Still exploring
Thinking about the same problem?
I'm always interested in what works, what fails, and how to test the difference.