A developer used Google Antigravity, an agentic coding platform, to build funsize-engineer, a Python package that displays contact info and ASCII art in the terminal. Instead of writing code manually, the developer acted as an architect, directing the agent to handle scaffolding, logic, and CI/CD setup. The package is now installable via PyPI.

Antigravity's core features include a Manager View for tracking project objectives and an Artifact-Driven Workflow. The agent generates implementation plans and code diffs for human review before execution. This Human-in-the-Loop model maintained control while offloading repetitive tasks like configuring GitHub Actions for dual publishing to TestPyPI and PyPI.

The process wasn't flawless. The agent solved a tricky versioning issue where setuptools-scm appended a `.dev` suffix to releases by forcing an explicit version environment variable. It also fixed a Mermaid.js diagram rendering problem on GitHub Pages by switching from markdown to raw HTML. These fixes demonstrate the agent's ability to troubleshoot complex, real-world engineering problems.