The Cognitive Boundaries Framework is a cross-field vocabulary for a problem that sits between disciplines: what happens when agents at the same scale combine, and why does the collective sometimes get smarter and sometimes get dumber than its parts?
The framework doesn’t compete with the formal tools in your field. It offers a small, precise vocabulary that lets you recognize the same structural pattern — in organizations, biological systems, AI deployments, and anywhere agents couple — using the same words. The payoff isn’t that the vocabulary is novel (most of the individual terms are inherited from existing literatures). The payoff is that the joint treatment lets you see connections between fields you wouldn’t think to compare.
The honest scope: This is a vocabulary layer, not a mathematical formalism. It complements mechanism-specific literatures rather than replacing them. The framework is strongest on agentic-system composition, cross-community knowledge failures, and structural pattern recognition across substrates. It’s weakest on paradigm shifts, ecosystems, and systems that don’t involve agents with all three boundaries.
This is also a work in progress. I’m developing it openly. Some of the entry points below are still being written — the ones that aren’t ready yet are marked. If your field isn’t listed and you think there’s a connection, I’d like to hear from you.
Enter through the door closest to your work
These pages are a work in progress
-
Active Inference and the Free Energy Principle — If you think in terms of Markov blankets, generative models, and variational inference. How the framework’s three-boundary decomposition relates to sensory, internal, and active states — and where the composition problem extends what active inference currently handles.
-
Basal Cognition and Scale-Free Agency — If you work on cognition in non-neural systems — bioelectricity, plant signaling, bacterial decision-making, slime mold computation. How the framework’s scale-free claim maps to competencies at every scale, and what it adds to the question of how local competencies compose into collective goal-directed behavior.
-
Extended and Distributed Cognition — If you work in the 4E cognition tradition — extended mind, distributed cognition, group cognition, cognitive scaffolding. How the framework’s Tool vocabulary and composition problem extend the insight that cognition isn’t confined to the skull, especially at the lateral-composition scale that Hutchins, Theiner, and others have been working on.
-
AI Safety and Alignment — If you work on multi-agent alignment, oversight, delegation, or boundaries of agency. How the framework’s structural claims — Cognitive-Causal independence, the Franz Ferdinand Effect, the supervisor-evaluation principle — generate specific diagnoses for alignment failures.
-
Organizational Theory and Management Science — If you work on bounded rationality, transactive memory, team cognition, or institutional design. How the framework systematizes organizational failure modes that the field has observed for decades, and what the three-boundary decomposition adds to the diagnostic toolkit.
The vocabulary
All terms are defined with precision at the Framework Vocabulary — with examples for both business and research audiences. The vocabulary is the common backbone behind all of the entry points above.
About the author
I’m not an academic. I’m a six-time founder with twenty-plus years of operating experience. I arrived at this framework by running companies and implementing systems in both production and research — I discovered the formal literatures afterward. The convergence between what I built from practice and what researchers arrived at independently from theory is what convinced me the pattern is real. More on the origin at The Shape of Composite Intelligence.
Contact
If any of this connects to your work — or if you think I’m wrong about how it connects — I’d like to hear from you. @rajjha on Twitter/X or raj@rajjha.com.