Overview
Dan Shapiro proposes a five-level progression model for AI-assisted programming, from basic autocomplete to fully autonomous software factories. Each level represents increasing AI autonomy and decreasing human involvement in the actual coding process.
The Breakdown
- Level 0-1: Spicy autocomplete and coding intern - Basic AI assistance where humans still write most code, using tools like GitHub Copilot for snippets and boilerplate with full human review
- Level 2-3: Junior developer and developer stages - AI handles more substantial coding tasks through pair programming, with humans transitioning from writing code to primarily reviewing AI-generated code
- Level 4: Engineering team management - Humans become engineering managers while AI agents handle most development work, collaborating on specifications and project management rather than coding
- Level 5: The dark software factory - Complete automation where humans never review AI code and focus entirely on system design and validation, with AI agents handling all implementation through extensive testing and tooling
- Real-world validation - Small teams are already operating at Level 5, building convincing software in months by designing systems that prove correctness through testing rather than code review