Bytery – a binary JSON protocol ~10x faster and ~10x smaller
Binary JSON with table reuse, but CBOR and MessagePack already own this space.
LLM token savings angle is interesting, but binary formats are crowded.
Developers optimizing API payloads and LLM context costs
MessagePack · Protocol Buffers · CBOR
I've been developing a data serialization format designed to be significantly lighter than JSON by eliminating repetitive key verbosity and optimizing the overall structure. Today, I'm open-sourcing the JavaScript SDK (lytok-js).
In our latest benchmarks using the WASM engine, Lytok achieves:
A 64.77% payload reduction.
50.19% faster network and parse speeds.
49.51% LLM token savings when feeding complex structures to AI models.
The core idea is that the structure is distilled through a definition map (which you can test in the Lab). This allows the engine to handle the heavy lifting so that thin clients and high-traffic networks aren't saturated with redundant data.
GitHub Repo: https://github.com/lytok/lytok-js
Interactive Lab: https://lytoklab.netlify.app/
I'd love to hear your thoughts, architectural critiques, or optimization suggestions.
Binary JSON with table reuse, but CBOR and MessagePack already own this space.
Another signed claim spec when JWT and Verifiable Credentials already exist.
The site sells a striking concept — "312M:1 transduction" and a 0xBF watermark on Solana — wrapped in a terminal-styled dashboard with a live calibration progress bar and infrastructure status (Protocol Core, Treasury, encrypted API, Ghost CLI). It looks great and leans hard into mystery, but the marketing claims arrive without documentation (the docs apparently 404) or technical proof on the page, so it reads more like an intriguing teaser than a convincing technical release.
Static Go binary coding agent that skips Docker and runs anywhere.
Finally handles circular refs and Maps where JSON and MessagePack both fail.
Custom notation cuts JSON size, but Protocol Buffers and MessagePack already dominate.