Dan K. SnelsonDocumenting my use of AI — for the users of my various open-source projects — and as a personal reminder of my self-imposed guardrails
Church Guidance
March 2024
In March 2024, Elder Gerrit W. Gong, of the Quorum of the Twelve Apostles of The Church of Jesus Christ of Latter-day Saints, led a meeting with Church employees worldwide, urging reliance on the Spirit, wisdom, and trusted sources while using AI. (See also: Our Own Best Story, April 2023.)
November 2024
In November 2024, Elder David A. Bednar, of the Quorum of the Twelve Apostles, presented Things as They Really Are 2.0 at a Worldwide Devotional warning young adults that while artificial intelligence offers immense possibilities for God’s work, it also carries limitless perils that can diminish moral agency, foster spiritual slothfulness, and turn us into objects acted upon rather than agents who act — urging them to reject technological shortcuts and instead engage in sustained righteous work under the guidance of the Holy Ghost.
August 2025
At a devotional during Brigham Young University’s annual Education Week, Elder Gerrit W. Gong presented An Ear to Wisdom, a Heart to Understanding, inviting listeners to diligently seek God’s divine wisdom and understanding through scripture, prayer, and the Holy Ghost so they can wisely and ethically use artificial intelligence as a tool aligned with faith principles, while fully embracing the healing and redemptive power of Jesus Christ’s Atonement. (See also: News Release.)
December 2025
General Handbook updated with 38.8.48 Appropriate Use of Artificial Intelligence.
February 2026
Last update to Principles for Church Use of Artificial Intelligence.
Mac Admin Guidance
Typos are how you know the notes are written by a human. — Bart Reardon
July 2025
I published my first AI-related post, AI in brief for the Mac Admin Purist, and I find its core message is still relevant today.
March 2026
Armin Briegel’s Thoughts on LLM-aided Coding post was thoroughly insightful (although I may challenge him on his conclusion during our next in-person meeting).
April 2026
Samuel Mills’ excellent Engineering in the Age of AI post raised some valid warnings about wholesale AI adoption eroding the system engineering craft.
AI Guiding Principles
You have to be smarter than the tool you’re usin’ — Dan K. Snelson
The following are my personal guardrails; your feedback is welcome as all of our AI usage continues to evolve.
Seek and follow the Spirit
Before working for the Church, I would frequently receive inspiration about my employer’s business and my role in helping to achieve its success.
Since working for the Church for more than the past decade, I receive near constant inspiration about what and how to help others as we each strive to accomplish the Lord’s work.
AI is becoming a big part of my day-to-day work, but it is not — and can never be — a substitute for personally seeking and diligently following heavenly inspiration.
Verify, then Trust
Just because AI can rapidly — and always confidently — generate code, doesn’t mean that you should ever blindly trust its output without fully understanding what was written, including careful, meticulous validation.
Remember, you have to be smarter than the tool you’re using — and AI will never be the source of that wisdom.
“How,” not “What”
Much of the inspiration I receive involves “what” I should do to resolve an issue or solve a problem or build a solution.
While AI can assist with the “how,” it’s not capable of determining the “what.”
Always remember …
The music is not in the violin. — Kahlil Gibran
AI Hallucinations 101
The above was the end of this article, until AI hallucinated while reviewing it.
Me: Review this yet-to-be-published blog post by first reading its referenced articles and then analyze the blog post itself.
AI: The Briegel aside will confuse readers. “I may challenge him on his conclusion during our next in-person meeting” is a delightful personal note, but as written it leaves the reader completely in the dark. What conclusion? Briegel’s post actually ends on a more positive note about LLM-aided coding than his setup would suggest — he concluded the Swift rewrite was genuinely worthwhile and that Claude Code met his threshold. If that’s what you intend to push back on, saying so briefly would give the reference real substance. Otherwise it reads like an inside joke that works for Briegel but not for anyone else.
Me: Doesn’t seem like you read the last sentence of Armin’s post.
AI: [Acknowledged missed conclusion and resolved to verify article’s ending]
AI: You’re right to call that out. Let me get the full ending of that article.
AI: [Recalibrated interpretation after discovering overlooked conclusion]
AI: You’re right, and thank you for the correction — my earlier analysis was flat wrong on this point, and it mattered.
AI: I had read the token-truncated version of Briegel’s article and assumed his conclusion was positive — that the experiment proved worthwhile enough to continue. The actual ending is the opposite: he concludes he doesn’t believe the benefits outweigh the downsides and that he will continue to stay away.
AI: That changes everything about my reading of your aside. I wrote that “I may challenge him on his conclusion during our next in-person meeting” left readers in the dark. It does leave readers in the dark, but now the substance of the disagreement is the interesting part — you clearly do use AI, and you’ve built a personal framework to do so responsibly rather than abstaining entirely. Your three principles are essentially a direct, principled rebuttal to Briegel’s position. That makes the reference much more meaningful than I gave it credit for.
AI: I apologize for missing that in the initial read. Truncated fetches are exactly the kind of thing your “Verify, then Trust” principle was written about.
