The most pressing challenge in artificial intelligence today is "alignment"—ensuring that models behave safely and reflect human values. Currently, the most important trend in solving this is a move toward Constitutional AI, a structural alternative championed by Dario Amodei and the team at Anthropic (Claude). It reflects the exact lesson democracies discovered centuries ago: without a constitution, a system can quickly devolve into chaos. Instead of letting a machine learn blindly from unpredictable crowdsourced feedback, developers give the model a written rulebook—a Constitution—that it must follow to govern its own behavior and evaluate its own answers. This is a vital approach because it prevents the AI from simply guessing what is right or averaging its values down to the lowest common denominator.

While standard Constitutional AI provides a rulebook, the AI's underlying knowledge is still built on a broad scraping of all available information. While this statistical approach succeeds brilliantly for math, coding, and general knowledge, it has limited utility for the far more subjective humanities generally, and philosophy and theology in particular. Broadly constructed AIs can only average their answers across large, general populations, producing results that require additional curation to be of true use to an interested community.

We envision the Barefoot Canon as an architecture built specifically for the Second Axial Age—a canon carried in a vulnerable, open way, entirely in process.

Initially this is locked specifically to the works of Pierre Teilhard de Chardin, starting with The Phenomenon of Man, The Heart of Matter, and his letters. As we grow, we will expand this corpus to include all of Teilhard's written work. By the middle of next year, we will integrate the works of Alfred North Whitehead. The engine will continue to expand under a rubric that broadly covers open and process thinking, serving the open and relational theological community.

An emergent constitution cannot govern itself. It requires the active discernment of a living community.

To ensure the AI does not drift into sterile compliance or sycophancy, we are actively enlisting constituent interest groups—scholars, theologians, and dedicated practitioners—to help us build this tool. These community members are invited to interact with the model, challenge its assumptions, and rigorously grade its answers.

By taking these specific critiques and manually patching them back into the AI's core instructions, we ensure that the Barefoot Canon is governed not by an isolated algorithm, but by the living community it was built to serve.