For most of history, changing what a large number of people believe was slow, expensive, and visible. You could see the newspaper, the pulpit, the campaign. Artificial intelligence has removed all three constraints at once. Belief can now be shaped quickly, cheaply, and without leaving a mark. The people who build these systems understand what they do. The people affected by them mostly don't. That distance has a name in our work: the persuasion asymmetry. The Institute exists to narrow it, not by telling people what to think, but by giving them the means to see how their thinking is being acted upon. The evidence is already unsettling. Nightingale and Farid (2022) found that AI-synthesised faces are rated as more trustworthy than real ones. When a fabricated signal outperforms a genuine one, trust stops being something earned and becomes something manufactured. We call that synthetic trust, and it is the central problem of the decade we're in.