Overview
A team used AI to rewrite the JSONata JSON expression language from scratch in Go, completing the project in just 7 hours. This case study demonstrates "vibe-porting" - using AI to recreate existing software in different languages by leveraging comprehensive test suites for validation.
Key Arguments
- AI can rapidly recreate complex software implementations when guided by existing test suites: The team built a working Go version of JSONata in just 7 hours using $400 worth of AI tokens, with the existing test suite serving as the validation framework
- 'Vibe-porting' is an emerging pattern for AI-assisted software development: This follows similar projects where AI recreates software functionality in different languages, enabled by comprehensive testing that validates behavioral equivalence
- Shadow deployments provide safe validation for AI-generated code in production: The team ran old and new versions in parallel for a week to confirm identical behavior before switching over
Implications
This represents a significant shift in how software can be ported and reimplemented. Teams can now recreate entire codebases in different languages within hours rather than months, dramatically reducing development costs and enabling rapid technology migrations when supported by robust test suites.
Counterpoints
- Heavy reliance on existing test coverage: This approach requires comprehensive test suites to validate AI output - projects without good tests would be much riskier to port this way
- Limited to well-understood, established software: Vibe-porting works best for mature software with clear specifications and expected behaviors, not cutting-edge or poorly documented systems