What it does: This tool solves my problem: You have a JSON file that changes regularly, and you want to track its history without storing dozens of full copies.
json-archive creates a .json.archive file next to your original JSON file. Each time you run the tool, it calculates only what changed and appends those deltas to the archive. You get complete history with minimal storage overhead.
The archive format is human-readable JSONL (not binary), designed to be hackable and easy to inspect, debug, and pipe into other scripts or web visualizations.
Repo: https://github.com/PeoplesGrocers/json-archive (AGPL so maybe don't look at it on work computers)
Why I'm posting: People keep saying AI coding makes them productive, but there aren't many concrete examples to study. So here's mine: - Bunch of working Rust with documenation
- Working tool with proper error handling, testing, and real examples
- Yes, some docs have that "AI feel" and the code takes a bit of a brute force approach.
- But I'm happy with the file format and the API it came up with
- It works and I only had to read/understand code rather than write everything from scratch
My question for HN: I have all the Claude Code and Cursor session transcripts that went into building this but I don't know a good way share them so you all can judge whether I'm using these tools effectively or like a noobIs there something out there? I'm looking for a way to publish these transcripts and maybe leave comments?
If you're curious about AI-assisted development or want concrete examples before adopting it yourself, I think this could be a useful data point. And if you know how I should share these sessions, I'd love to hear suggestions!
mutant•1h ago