Forgotten #simplisity . one of my colleagues long time ago show me #forth . It was stack based language with #factorisation principle. With every word you program gets better and better and execution mechanisms was super simple but powerful. You get compiler , interpreter and os in one shot. #forth really help to develop complex systems . In a few years I fail in love with #lisp . It was more complex but as same as #forth was build around simple but single data structure - list . #lisp was homophonic language that was self aware and capable to change itself. You still could find a quote from Allan key about most butifful program - lisp interpreter in lisp. #smalltalk was another language that fight for simplicity that cover what lisp was not able to do . Simple objects that send messages and turn your software to a soft internet. #self lift this idea even more to extreamly simple model with out class/objects separation. Smalltalk and self was not homoiconic but it was full reflective . According to a many researches smalltalk has the lowest cognitive load and the highest #DX in ide. Last simplicity guru - #prolog - based on idea of horn clause and first order logic. You have terms and few logical operation together with homoiconisity and ability to write meta interpreters. It is just blow my minds. In modern world we still widely use #datalog as simpler and limited data oriented language. Lisp , prolog was designed for #ai and reasoning . Smalltalk was a first language that master #agents . All of this languages is a masterpiece of simplicity and has low cognitive load and minimal syntax and features. All of them build around of internal dsls and code that reflect domain with out language noise.
#lisp and #prolog was tailored for #ai challenge . How #prolog could help with #knowledgegraphs #hypergraphs and multidimensional #ontology ?
I try to connect few dots - #multimodal #databases , #relational #graphs and #commonlogic and #datalog and #prolog for #onlologes and constrains . Some times your ontology is dynamic and derived from data patterns . I keen to teach #llms to do a bit of #prolog . More I play with schema for graphs more I see a lot of work for solvers .
#crdt for #licalfirst and #sovereign data
#Wallets , #agent and #vaults require personal and decentralized persistent... #ssi #web5
we need #agents not #wallets
https://volodymyrpavlyshyn.medium.com/personal-knowledge-graphs-semantic-entity-persistence-in-relational-model-d5692bb8e8bb I model graphs with string labels and hypergraphs in a regular relational DB in a few articles. It makes Personal Knowledge Graphs more portable for different kinds of applications. How to model #heterogeneous #graphs of complex #entities as nodes in a plain relational and portable #databases ? Lets discover options together #eav #json #documentstore
#hypergraphs #knowledgegraphs
#AI models need a #hypergraphs to represent a complex multimodal knowledge
https://volodymyrpavlyshyn.medium.com/personal-knowledge-graphs-in-relational-model-39528f1ee45b So, go from #directedgraph to #hypergraph on steroids. How does #namedgraph help us simplify structure? When using a different kind of graph. Yep, all this is a pure relational model that works better with a #datalog. Relational databases could be a good choice for #personalknowledge #graphs and #ai #empowered applications