[conzept]

I think the future holds a synthesis of LLM functions with semantic entities and logic from knowledge graphs (this is called "neuro-symbolic AI"), so each topic/object can have a clear context, upon which you can start prompting the AI for the preferred action/intention.

Already implemented in part on my Conzept Encyclopedia project (using OpenAI): https://conze.pt/explore/%22Neuro-symbolic%20AI%22?l=en&ds=r...

Something like this is much easier done using the semantic web (3D interactive occurence map for an organism): https://conze.pt/explore/Trogon?l=en&ds=reference&t=link&bat...

On Conzept one or more bookmarks you create, can be used in various LLM functions. One of the next steps is to integrate a local WebGPU-based frontend LLM, and see what 'free' prompting can unlock.

JSON-LD is also created dynamically for each topic, based on Wikidata data, to set the page metadata.