TL;DR: Internet Object (IO) is a lean, schema-first, JSON-compatible format that cuts structural noise, improves clarity for modern systems, and reduces tokens by ~40–50%.

Playground: https://play.internetobject.org

---

I started exploring this idea in 2017 after repeatedly running into the same pain points with JSON while building distributed systems and structured data pipelines. Rather than extending JSON or adding more layers on top of it, I wanted a format that was clean, schema-first, human-friendly, and still compatible, where that compatibility matters.

The concept evolved slowly over the years, with several redesigns, dead ends, and restarts - until it eventually converged into what I now call Internet Object (IO). The story behind this evolution is here:

https://internetobject.org/the-story/

Although IO was not created with LLMs in mind, its structure ends up being significantly more token-efficient (around 40-50% fewer tokens compared to JSON), which has become a practical advantage in today’s workloads.

I've written a practical guide showing how JSON developers can transition to IO, with syntax explanations: https://blog.maniartech.com/from-json-to-internet-object-a-l...

For side-by-side comparisons with JSON, see the following link:

https://www.internetobject.org/io-vs-json/

There is also an interactive playground if you'd like to try the format directly:

https://play.internetobject.org

https://play.internetobject.org/simple-collection

This is a soft launch to gather early feedback - I would appreciate any thoughts from the community.