The story

RSS worked. Why not for AI?

I'm Raphaël Reck, an IT Systems & Software Consultant based in Sophia Antipolis, France, with over 20 years in the field. I run Roots Time Corporation.

LLMR started as a practical problem. I was building with AI agents and watching them burn 30,000 tokens just to understand a 13-page website; most of it wasted on HTML navigation, footer links, and script tags that added zero information. RSS solved this for feed readers in 2000. There was no equivalent for AI systems.

So I built one. The full benchmark, measured on my own site, is documented on my blog:

→ I measured how many tokens an AI needs to understand my site (raphaelreck.com)

The numbers

Measured on raphaelreck.com. 13 pages. Not estimated. Actual.

HTML (all 13 pages) ~30,000 tokens

LLMR index ~3,300 tokens

Reduction 89%

HTTP requests 13 → 1

What I do

Beyond LLMR, I consult on systems architecture, legacy modernisation (Drupal, Laravel), API integration, and AI tooling. If you're building something that needs to talk to AI systems efficiently, I'm available.