Show HN: Lytok – A binary format with 44% fewer tokens than JSON

2 points

4 months ago

I’ve spent the last year building Lytok, a binary data notation format born out of a personal obsession with data density and infrastructure efficiency.

While JSON is the standard, its verbosity becomes a bottleneck in high-throughput systems and AI pipelines. Lytok focuses on four key pillars:

Token Efficiency: Up to 44.80% fewer tokens than JSON, significantly reducing LLM inference costs.

Storage Density: Up to 60% reduction in disk space usage.

Bandwidth Economy: Smaller payloads for faster, cheaper network transfers.

Strong Typing: Native support for complex data types without the overhead.

This is a project of passion I’ve been refining since last year. I’ve built a web-based Lab so you can test these claims with your own datasets.

Check out the Lab: https://lytoklab.netlify.app/ I'd love to hear your technical feedback.