Incremental Implementation

Incremental Implementation

Overview

Build in small, complete slices rather than large unfinished layers. Each slice should deliver a thin but working piece of value and be verifiable.

When to Use

• Implementing any non-trivial feature
• When previous large attempts have stalled or produced hard-to-debug code
• When working with an agent that benefits from frequent feedback loops

Core Process

1. Identify the thinnest useful slice that touches the full path (UI → logic → data, or API → service → storage).
2. Implement only that slice.
3. Verify it works (tests + manual/runtime check).
4. Expand to the next slice.
5. Refactor as needed while staying green.

Principles

• Prefer vertical slices over horizontal layers early on
• Keep the system in a working state as much as possible
• Make the first slice intentionally small to validate direction
• Resist adding “while I’m here” improvements outside the current slice

Anti-Patterns

• Building all the models first, then all the services, then the UI
• Large uncommitted or untested changes
• Scope expansion during a slice

Verification

After each slice:

• Relevant tests pass
• The slice is demonstrably working end-to-end
• No obvious regressions in existing behavior