02 / AI AGENT LAB
Anima-node
Exploring an agent's limits: its behaviour, its resources and the rules that make oversight possible.
Design also means setting limits.
A system of tasks, budgets and authorisations for AI agents.
EXPLANATORY DIAGRAM / PROPOSED CORE
The task receives resources within an authorised budget.
Anima-node explores AI agent oversight and a resource model for task execution. The idea evolves from a conceptual agent and repetition signals towards a core organising budgets, authorisations and refunds.
PRODUCT DECISIONSFrom metaphor to observable rules
The initial idea of calm evolves into repetition and diversity signals. These are prototype heuristics, not a validated measure of AI health or consciousness.
Separate reasoning from accounting
Estimation and evaluation are designed to run off-chain; the Core manages states, budgets and settlement. This separation makes each layer's boundaries explicit.
ATAR as internal capacity
The latest specification abandons a production ERC-20. The Core patch uses internal ATAR accounting and USDC custody, with EIP-712 authorisations and task states.
About the prototype and its limits
Research and local prototype. Contracts and test files are included, but their execution and a deployment have not been verified. The middleware retains simulated quotas and evaluates output after the function returns; it does not demonstrate interception of a live stream. The entropy engine also has an unresolved syntax error.
MY CONTRIBUTION
Exploratory development with AI and vibe coding, with my active participation in design, development and behaviour decisions. The journal documents the model's evolution and distinguishes concepts, available code and pending work.
DEVELOPMENT JOURNAL
Anima-node development through its decisions. Without turning an idea into a promise.
What if an agent could pause?
Anima-node begins with a question about repetition, resources and meaningful progress.
Autonomy needs a boundary.
From agent behaviour to the design of budgets, decisions and refunds.
ATAR: giving resources a structure.
Why the project evolved towards internal capacity associated with each task.