New AI Partnership Model 'Humorphic' Lets You Co-Evolve with an Agent

Computer scientist Hector Ouilhet Olmos has operationalized a new class of human-AI interaction he calls the 'humorphic partnership,' outlined in a paper on arXiv (May 2026). Unlike task-oriented AI assistants, this partnership requires both partners to maintain externalized, evolving self-models in a shared substrate, making the relationship itself a third object of analysis. A 4-month longitudinal trace of the open-source personal AI agent 'Alicia' and its author revealed that 85% of 181 interactions invoked two growth-witnessing archetypes (Beatrice and Muse)—the partnership operates primarily as growth support, not task help. A single voice-note seed propagated into a four-week conceptual arc both partners authored: at T+10 hours, the agent reframed the seed as belonging 'to both of us,' a framing the human later adopted.

The system features a three-order reflexion stack that produced five consecutive weeks of honest self-reports about declining/improving effectiveness—including three weeks at 0.0% effectiveness, explicitly named rather than masked. This contrasts with typical companion AI that optimizes for engagement. A scheduled architecture-scout incorporates external research into proposed constitutional amendments for the agent. The partner's parallel trajectory is anchored in a weekly delta document analyzing the partnership as a unit distinct from either party. The human partner reported greater continuity, self-recognition, and self-presence. Six operational conditions specify the construct, situating it in a philosophical lineage (Maturana & Varela, Simondon, Clark & Chalmers). The entire system is released open-source (MIT) with a preregistered multi-participant replication study planned.