We Rewrote JSONata with AI in a Day, Saved $500K/Year
reco.ai
March 26, 2026
8 min read
🔥🔥🔥🔥🔥
61/100
Summary
Reco utilized AI to rewrite JSONata in one day, resulting in a cost savings of $500,000 per year. The process involved using an existing specification and test suite to guide the AI in implementing the code until all tests passed.
Key Takeaways
- Reco developed gnata, a pure-Go implementation of JSONata 2.x, in seven hours using AI, resulting in a $500K annual cost savings.
- The previous JSONata implementation in JavaScript incurred approximately $300K per year in compute costs due to RPC overhead.
- The new gnata implementation eliminates the need for RPC calls, reducing latency and improving performance for evaluating JSON expressions.
- The project involved porting the official JSONata test suite to Go and achieving 1,778 passing test cases with a total token cost of $400.
Coverage & discussions
Hacker News · reco.ai
Community discussions
Hacker News discussion ↗
Historical count: 5 · 5 saved in sample · Historical sample
Analysis ready
Community Sentiment
Positives
- The use of AI to rewrite JSONata in Go significantly reduced operational costs, saving approximately $500K annually, which highlights the potential for AI to optimize software development.
- Transitioning from a JavaScript-based architecture to Go not only improved performance but also streamlined the deployment process, indicating a successful application of AI in enhancing system efficiency.
Concerns
- Concerns arise about the long-term maintenance of the AI-translated code, especially given the existing open issues and pull requests, which could impact its reliability and sustainability.
- The high operational costs of $300K per year for running JSONata in a Node.js environment raise questions about the efficiency of their previous architecture and whether AI was truly necessary for the rewrite.
