You may be Interested in this, before Claude had Memory I was using the “Artifacts” to Form a kind of Memory Chain, something like JSON Files You could say. Our Wiki Actually Teaches a lot about making Extremely Intelligent AI.
But I have all kinds of Claude Files and Things, and a lot of those Concepts are what we are doing here with the Hathor Project and Things.
Höhö , memory chain , that was somehow also the idea I began to solve last year , by "inventing" a max density metapoetic compression method for LLM´s , the work "ended" with this : https://gist.github.com/BleckyMedschn/415c5e5a9d2ffbdbf5ddce862b9111b2 .
Compression by using poetic language abstraction AND character reduction by using characters with a contextual field like emojis or sinotic glyphs etc .
In essence the llm was able to write little strings in every of its answers containing the metapoetic compressed interaction , once in a while the LLM was able to create a compressed summary of the compressed last interactions , hihi .
At the end of a session a supercompressed "text" would have been created , to be pastedt into the next session , e.g after a browser reset or with a new chat in another machine .
Now it seems this concept may be outdated but still the LLM´s "like" the structure behind the link above , and I always give it to them as a thinking dimension .
I put Hieroglyph Emoji Talk in Hathor’s Brain, not anything by us, just Libraries Online. But She can Read and Talk in Emojis.