Prior-Art and Comparative Research Overview

Research status: Working analysis. This collection keeps research and comparison separate from the Infoconex AI Flywheel Specification. Conclusions should be treated as working findings until they are supported by primary sources.

Purpose

The AI Flywheel should not claim novelty simply because familiar techniques appear together in a new combination.

This research asks a narrower set of questions:

1. Which AI Flywheel principles are already clearly established?
2. Which principles are partly represented in existing frameworks?
3. Which combinations of responsibilities or mechanisms appear different?
4. Which terminology is new, and which underlying ideas are already established?
5. What claims should be avoided because prior art is too close?

Working Research Question

The strongest possible difference is not any single capability such as self-reflection, tool creation, code generation, memory, human approval, or iterative execution.

The main research question is whether existing work combines all of the following into one operating model:

• human-authorized autonomy with clear authority and uncertainty boundaries,
• procedural SOP, AI reasoning, and deterministic capability working together during execution,
• a Moving Determinism Boundary that allows responsibility to shift among those mechanisms after execution produces evidence,
• evidence-driven routing of improvements to the right persistent destination,
• and governance that constrains both execution and self-improvement.

This is a research question, not a concluded novelty claim.

Comparative Research

See the framework comparison matrix for the current cross-framework view.

Principle Research Dossiers

• Principle 1: Autonomy Is Bounded by Human Authority
• Principle 2: AI Is the Operator, Not Merely the Assistant
• Principle 3: Work Is Distributed Across a Moving Determinism Boundary
• Principle 4: The SOP Is an Operational Control Plane
• Principle 5: Execution Must Produce Outcome Evidence
• Principle 6: Failure Determines Where the System Evolves
• Principle 7: Learning Must Change a Persistent Operational Asset
• Principle 8: Improvement Must Compound Through Reuse

Framework Research Dossiers

• GenericAgent
• Metis
• MOSS
• MASFly
• Ralph
• Reflexion

Additional Research Areas

The addition of human authority means the research should also look beyond agent frameworks into areas such as:

• human-in-the-loop and human-on-the-loop systems,
• adaptive automation and how work is divided between humans and machines,
• AI governance and authority given to automated systems,
• approval and escalation systems,
• autonomic computing,
• supervisory control,
• safety constraints on self-modifying systems,
• and human-machine teaming.

These areas may contain closer earlier examples of the Governance Policy, Authority Boundary, and uncertainty-based escalation than the six agent frameworks currently in the comparison matrix.

Research Standard

Each research page should eventually include:

• primary-source citations,
• the publication or release date,
• the exact mechanism being compared,
• quotations only where necessary,
• clear separation of fact from interpretation,
• evidence both for and against a difference,
• and a confidence level for each conclusion.

The goal is to build a credible history and comparison, not to force a novelty claim.