If you're feeding structured data to LLMs, you're probably wasting tokens on JSON overhead: repeated keys, quotes, braces, commas. I built LEAN (LLM-Efficient Adaptive Notation) to fix that. Benchmark results (avg token savings vs JSON compact, 12 datasets):
Detailed Analysis
Detailed analysis coming soon.
Read original article →