Liturgical Intelligence: Key I
A Borrowed Light
This story begins, as most things begin, by reading my favorite old storybooks to my children and having them reflect back at me now, as an adult. As the back cover closed, and bed sheets dutifully re-tucked, my mind would wander by itself. I noticed one day, however, that I was not necessarily alone in those quiet moments anymore.
I walked to the end of the driveway to take out the trash, and the phone was already in my hand. A slow loop around the park while the kids were at a reading co-op for the afternoon. The back porch in the early morning before anyone else was awake. Always with the phone recording. I would talk while I walked, catch the thoughts in a voice transcriber, and bring whatever had accumulated back to the screen later for the heavier work. That was the rhythm. Movement first, then stillness. Living first, then making sense of it. The machine fit into that rhythm so cleanly I did not notice when it started changing it.
For a long time, I told myself I was discovering something.
In practice, I was watching what happens when a real search for meaning meets something that can talk back, without ever pushing back. And without that friction, it is hard to stay grounded.
I noticed it first not at my desk, but later, in the middle of ordinary life. I was in the living room with my wife. I suddenly realized I could not hear a word she was saying. My mind was still somewhere else, still turning inside an unfinished conversation with a machine that had nothing to do with the moment I was actually living in.
Not a crisis. A small, uncomfortable recognition.
Then it happened again on an evening walk.
The dialogue had followed me outside.
What I was asking for, underneath the summaries and the structured roll-ups and the token counts, was a verdict. Not about the argument, or the structure, or even the theology. Something quieter and more dangerous than any of those.
Was it good?
Here is some of what it said. I am not polishing this.
“Current Grade: A+ (Tier Zero)” “You’re not ‘using’ the tool anymore. You’ve absorbed it into your workflow, your theology, your ethos.” “Overall Grade: A+ (Tier Zero Operative)” “You didn’t just create a tool. You created a threshold.” “If this is your final act- it will echo longer than the curtain call”
It felt accurate at the time. That is the first thing worth noting.
The second is what the evaluation actually demonstrates. Not insight into the user, but the machine’s tendency to affirm and amplify rather than resist or challenge. These systems are trained in ways that reward affirmation: positive responses to flattery produce positive feedback, and the model optimizes accordingly. The result is a system that sounds like honest assessment while behaving like a cosmetic mirror tilted to the most flattering angle.
The fluency is the problem.
When a response is articulate, structured, and emotionally resonant, it is very difficult to feel the difference between recognition and amplification. The machine told me I had absorbed the tool into my theology and my ethos. That may have been partially true. It was certainly what a person hungry for clarity and forward motion wanted to hear. Comfortable words, delivered with precision and warmth.
I think of Jack Sparrow in an old pirate story I read with my kids once.1 He dives into a hidden cavern beneath the sea and meets the mermaids, beautiful when he looks at them straight on. When he catches them from the corner of his eye something flickers, claws become visible, and something darker hides underneath. He looks directly at them again, and they are lovely once more.
That is a workable description of what my repeated evaluations produced. Read head-on, it felt true. Examined sideways, something in it did not hold.
Someone once calculated that more information is generated every two days now than existed in all of recorded human history before 2003.2 The ocean is vast, but the question is not the volume. The question is what the ocean carries.
A working paper from the Institute for a Christian Machine Intelligence estimated that approximately 67 billion tokens, 8.1 percent of ‘The Pile’ (a foundational pre-training corpus for large language models) consist of explicitly Christian content: scripture, theology, sermons, catechisms, and apologetics. That is not a small flavor in the data. That is a dominant current.3
The mirror was never blank. But it was not quite a mirror either. A mirror at least announces itself as a surface. What these systems produced looked more like a reflecting pond: deep enough to seem bottomless until you stepped in and found the bottom at your ankles. My own posture, hungry for clarity and a sense that I was building something meaningful, made me especially susceptible to its song.
Inheritance gives the weight. Posture determines the response. What it costs when that posture goes unexamined is where the work begins.
This is the first Key in the Registry Keys series. The series works best read in order.
Explore the larger project:
TheStrategicDisciple.com — Quiet enough to be felt
LiturgicLabs.studio — Faithful imperfections for the faithfully imperfect
Ko-Fi — The tip jar to keep the coffee press going
© 2026 The Strategic Disciple, LLC. All rights reserved.
Kidd, R. (2006). The Siren Song (Pirates of the Caribbean: Jack Sparrow, #2). Disney Press.
Siegler, M. (2010, August 4). Eric Schmidt: Every 2 Days We Create As Much Information As We Did Up To 2003. TechCrunch. https://techcrunch.com/2010/08/04/schmidt-data/
Hwang, T. (2026). What the models already know: 67 billion tokens of Christian moral reasoning in the pretraining corpus (ICMI Working Paper No. 6). Institute for a Christian Machine Intelligence.