People and agents, one workspace.
Appkask-Studio is my research and development project. It studies how people and agents work together on shared files and processes. People approve, and every step remains traceable.
Six layers, from the files up.
Each layer builds on the one below. Every step is traceable, and people approve.
Six layers, from the files up
Each layer builds on the one below. People approve at every layer, and every step leaves a record.
Authority sits on the step.
The question behind Appkask-Studio comes from my work on agent-based modelling: how many actors act under shared rules, and how their actions can be verified afterwards. In Appkask-Studio, a BPMN process model becomes an executable net. Human tasks and agent tasks are the same kind of step, and approval is part of the model. Two working papers are in preparation: "Authority on the step, not the lane: executable semantics for human–agentic BPMN" and "The Map is the ontology: a harvestable type system for agentic workspaces". The research notes below document the work as it progresses.
From the research log
Notes written during the work.
Agentic tasks and lanes in BPMN
How a process diagram shows which tasks agents take on, how their results are checked and where a person approves.
Status cascade on dependency networks
How a change of status travels through a network of dependencies, and what that means for planning.
A digital twin updated by the status cascade
A model of operations that stays current because every status change updates it.
From SEO to AXO
Websites that agents can read: a guide file, a clean structure and a query interface.
See it at work.
aipokit.com shows use cases and prototypes from the project in detail.