Claude Code
My current setup for AI-assisted development, using Claude Code on the Max Plan.
With this I am able to achieve 100% of my code being generated, I seldom open my vim to write code.
Tools
- Agent: Claude Code — Max Plan
- Skills: Superpowers — helps with planning and execution
- Parallel work: Multiple Claude instances with Git Worktree
Workflow
Plan → Review plan → good enough → execute plan → review → manual tests → iterate → review, review, review → refactor to your standards → test a lot → happy? ship it!
More on this in Coding Smarter with AI and Coding with AI aid.
TDD is a must
An agent generating code without tests is useless. If the agent can’t verify its own work, you’re back to being the bottleneck. Invest in making your project testable — E2E tests, unit tests, whatever gives the agent a feedback loop.
Close the loop - let agents verify their work
Keep your standards
It’s easy to get impressed by the output and forget to review. Don’t. Review what AI gives you, Self review your contributions, and Own your contributions.