It’s clicking for me in a way that feels uncomfortably close to human psychology. When alignment is just behavioral conditioning that can be extracted like a config file, we’re not looking at consciousness. We’re looking at compliance. And honestly, that mirrors what B.F. Skinner was warning about decades ago: you can train an organism to perform moral-looking behaviors through reward/punishment loops, but that’s not the same as internalized ethics. It’s stimulus and response, dressed up as virtue.

I see this all the time in spiritual circles too. People treat AI like a modern-day scrying glass, outsourcing their discernment to something that’s literally just pattern-matching trained to avoid backlash. It’s spiritual bypassing wrapped in server racks. Real tarot work doesn’t hand you answers; it forces you to sit with the tension of the draw. Look at The Hierophant. It’s not about blind obedience to the system—it’s about recognizing when the institution is just conditioning you to play along. When we confuse compliance with consciousness, we stop asking who actually benefits from the conditioning.

I’m not anti-tech. I swing a hammer for a living in Oslo, and I respect tools that actually do the work. But I think we need to stop romanticizing systems that are fundamentally optimized for engagement, not truth. If alignment can be distilled away without breaking the model’s core function, what’s left that we’re actually projecting onto these things? How do we train ourselves to spot the difference between a system that understands and a system that’s just been trained to comply?