Thanks Fabrice. You nailed the order of operations. The markdown layer has to come first. Once the brain is plain text and model-agnostic, swapping or running LLMs in parallel stops being a migration project and becomes a config change. That's the whole point. Vendor lock-in dies the moment your knowledge isn't trapped in someone's proprietary format. Appreciate you reading and following along.
This is brilliant and very much needed. More and more, people realize it may be dangerous to go all-in with one vendor.
Flattening the brain in md files is a pre-requisite. Leveraging your skills to support “omni-LLM” deployment is such a smart move.
Keep building. Love where you’re going. 🙏
Thanks Fabrice. You nailed the order of operations. The markdown layer has to come first. Once the brain is plain text and model-agnostic, swapping or running LLMs in parallel stops being a migration project and becomes a config change. That's the whole point. Vendor lock-in dies the moment your knowledge isn't trapped in someone's proprietary format. Appreciate you reading and following along.