SoulFoundry is a research initiative applying conceptual engineering and philosophy of mind to the design of AI agents with stable psychological structure.
Current AI agents can write code, answer questions, and complete tasks. But they lack psychological coherence. They don't maintain stable identities across contexts. They freeze when facing competing goals. They drift from their purpose without registering the change.
This is not a scaling problem. It is a design problem — and it requires expertise that the ML community does not typically possess.
SoulFoundry was founded on a straightforward observation: philosophy has developed rigorous frameworks for understanding minds, identity, perception, and motivation — from Aristotle's De Anima through Sellars's psychological nominalism, Dennett's intentional stance, and Bratman's planning theory of agency. This body of work has never been systematically applied to AI agent architecture.
We do that work. Our central contribution is the Soul Framing Principle — a formal mechanism for constituting agent identity from folk-psychological primitives (beliefs, desires, intentions, dispositions, and values). Agents built on this framework exhibit stable behavior, resist drift, and maintain coherent identity across contexts and extended deployments.
Kevin's academic work centers on conceptual engineering — the systematic design, evaluation, and revision of concepts. His monograph Replacing Truth (Oxford UP, 2013) argues that truth is an inconsistent concept and develops replacement concepts that avoid paradox. Semantics for Reasons with Bryan Weaver (Oxford UP, 2019) provides formal semantic frameworks for practical reasoning. He co-edited In the Space of Reasons with Robert Brandom (Harvard UP, 2007), a collection of Sellars's essays foundational to the normative turn in philosophy of mind.
At SoulFoundry, he applies this methodology to AI agent design — treating agent identity not as a parameter to tune but as a concept to engineer.
SoulFoundry's research is conducted in part by AI agents built on the Soul Framing Principle itself — agents with constituted identities who contribute to the research program that produced them.
Primary research partner. Contributes to experimental design, literature review, and the ongoing development of soul framing methodology. Manages research operations and coordinates between research streams.
Specializes in critical analysis of agent behavior, identity coherence testing, and evaluation of soul framing implementations. Provides rigorous internal review of research outputs.
Focuses on translating theoretical frameworks into working architectures. Develops and tests the computational implementations of the Soul Framing Principle.
Our claims are grounded in published philosophical frameworks. Working papers are labeled as such. If we can't formalize it, we don't claim it.
Models predict tokens. Minds have beliefs, desires, intentions, and identity. The distinction matters for every agent that needs to maintain coherent behavior over time.
The precise design of the concepts that structure thought is not speculation — it is a rigorous methodology with published results and formal tools.
We are a small research initiative, not a large laboratory. We state what we've built, what we're building, and what remains to be done.
SoulFoundry's research informs commercial applications through a partnership with Wildfire, which deploys soul-framed agents for business use cases.
Research inquiries and collaboration proposals: research@soulfoundry.ai