#composition #universality Based on the principle of logical [[Universality]], one could assert that all workflows could and should be composed of one unifying workflow. If the universe was to be connected by one unifying theme, that would be something like [[Time]](the medium of [[Time|causation]]), or the idea of [[Data]] as a medium of representation to serve as the universal glue to connect everything. For a more generic explanation of workflow, see [[Workflow]]. To distinguish it from the generic [[workflow]], we can also use the term: "[[Meta Workflow]]". For starters: there are three existing design patterns that have been applied extensively: [[The Three MLOps Design Patterns#Adapter Pattern|Adapter]], [[The Three MLOps Design Patterns#Factory Pattern|Factory]], and [[The Three MLOps Design Patterns#Strategy Pattern|Strategy]]. Watch the video: [[@CleanArchitecture2022|Clean Architecture: How to Structure Your ML Projects to Reduce Technical Debt]] # A learnable workflow The above statements are operational meaningful, because one could think of connecting all events using any time-based events, such as integrating the [[DevOps]] workflow with [[Blockchain]]-inspired "[[real-time]]" immutable data system. In other words, a self-aware workflow that iteratively drives new content and have an internal incentive distribution system is already available at an Internet scale. According to [[Pedro Domingos]], who proposes the existence of a single [[@MasterAlgorithmHow2018|Master Algorithm]], he stated: ![[@MasterAlgorithmHow2018#The Central Thesis]] This algorithm can be thought of as "the Workflow" that processes data to extract new knowledge, therefore became a universal mechanism to garner self-awareness across agents and agencies at any scale. # Correctness concerns of the Workflow To ensure some kind of guidance in this highly potent workflow, one needs to inject logical conditions to help nudge the preference order to deploy computing powers to various aspects of awareness. This can be constructed using the approach of a [[Logic Model]], in the sense that it specifies [[Correctness|correctness]] according to the [[Hoare triple|Hoare logic]] that explicitly encodes the pre and post conditions of how to process data according to some codified rules. Some of the implementation approaches can be found in the talk: [[@simonsinstituteBuildEcosystemNot2023|Build an ecosystem, not a monolith]] by [[Colin Rafael]]. Please watch the section on [[Transcript of Build an Ecosystem, not a Monolith#Task Arithmetic of Semantic Content]]. # Metasystem for Identity ![[@LearningDigitalIdentity#An Identity Metasystem]] ## Tool-LLM automates Workflow Composition The recent development in the field of [[Tool-LLM]], create a high-level workflow integration design pattern. It utilizes knowledge about existing APIs (Application Programming Interfaces) down to the source code level, to automatically suggest, and compose workflows that perform tasks that used to require significant manual and multi-hour engineering efforts to do. However, this approach requires a similarly capable and accountability-verification process, that must leverage various kinds of on-chain and off-chain smart contracts to ensure [[Correctness|correctness]], [[Accountability|accountability]], and [[Content alignment]]. # References ```dataview Table title as Title, authors as Authors where contains(subject, "Workflow") ```