Editor's note (added August 10, 2026): The "guardrails" I described below have evolved since this post. Part 2 covers some of that shift, and the current, more accurate version lives on the About page.
It’s just before midnight, January 5, 2025. My wife Robbie and I just got home from a post-holiday dinner party with some Northfield friends. I’m in bed reading the exciting finish to the latest thriller in the Orphan X series while she’s stuffing her suitcase, preparing for a 4 am departure to the airport for her women’s group annual winter getaway to Florida.
Suddenly, my heart goes apeshit (my term for AFib’s irregular racing heartbeat), which hasn’t happened since I was diagnosed three years ago. I privately assume that since I don’t have any pain, it’ll stop in a couple of minutes. I don’t want to tell Robbie, because she’ll freak out and might bail on her trip, which she badly deserves.
We fall asleep, and when our alarms go off at 4 am, my AFib is still rocking. I drop her off at a friend’s house, kiss her goodbye for a week, saying nothing about my ticker. I head back home for a few more hours of shut-eye, hoping for a cure. No such luck. Time for a chat with Claude.ai. Questionable call, I know. Should I go to the ER? The full story is coming up in Part 2 of this series.
What this AI series is about
I’ve been using AI seriously for two years, a mix of personal, business, and civic. It’s helped me think more clearly about my heart condition and recovery, analyze city council meetings, research data centers, guide my Northfield.org strategy, and improve my mountain biking skills. It's also made stuff up, suggested some stupid ideas, and once created a Valentine's card for my wife that repeatedly misspelled everything.
So it’s very useful, but not trustworthy by default.
I also don't want AI to be something that happens to us. I want to figure out how regular people can use it to be better citizens: to understand a city budget, parse a police policy, or show up to a council or school board meeting with sharper questions.
I’m not done figuring this out. I’m definitely mid-process. You’re likely to see future posts with “In Part 13 I said… but I’ve since discovered that it’s more likely that...”
How I got here
I started using AI in December, 2022, about a month after ChatGPT launched. Initially, I treated AI like a newer version of Google until April 2024 when I read Ethan Mollick’s book Co-Intelligence: Living and Working with AI, and started following his Substack, One Useful Thing.
He advised immersion: don't just read about it. Use it for everything you can, for 10 hours a week. That inspired me to start the Northfield AI User Group (NAIUG) in early 2024, but I shut it down after two face-to-face meetings due to poor attendance. I remained a wary AI enthusiast.
My first big AI disappointment
ME: Create a Valentine’s card for my wife, Robbie. We’ve been married 50 years. Include a background of a sunset at sea. Make the card whimsical.
ChatGPT/Dall-E: Here’s your whimsical Valentine’s card for Robbie, celebrating 50 years of marriage with a sunset at sea theme. I hope it captures the essence of your enduring love and the romantic spirit of this special occasion. Would you like any adjustments to this design, or is there anything else you’d like to add to make it perfect for Robbie?
ME: Why is everything misspelled?
After 30 minutes, I gave up and bought her some flowers from Forget-Me-Not florist in downtown Northfield.
My first AI eye-opener (the moment I realized AI was different)
Ethan Mollick’s book got me interested in using AI as a thought partner. He wrote:
AI can assist us as a thinking companion to improve our own decision-making, helping us reflect on our own choices (rather than simply relying on the AI to make choices for us). We are in a world where human decision-making skills can be easily augmented in a new way.
I asked Claude about this role:
Humans are still superior thought partners in terms of understanding nuance and context, relating concepts to the real world, thinking creatively, providing emotional awareness, challenging assumptions, having unscripted conversations, and simplifying complex ideas.
His (hah) response surprised me, as I was expecting a one-sided treatise on all the advantages that AI had over humans. Was he being honest? Could I trust him to be self-critical? Do those anthropomorphic questions even make sense?
It sent me down the road of using AI to augment my thinking, and that has continued ever since. I’ve learned how I can suppress its sycophantic tendencies, crank up its bullshit detector, engage with my dad jokes, and shut up when I don’t need any more prompting.
Guardrails: my AI rules of engagement
AI never decides what I think about a person or an issue.
Whenever I start working on a first draft for a post, I don’t use AI. Writing is thinking, so I don’t want to outsource that to AI.
I always use AI as if I had hired an experienced editor to critique my first draft. I ask it to identify holes in my thinking, suggest better overall organization, look for ways to simplify, and generate options for better titles and subtitles.
I often ask two AI chatbots to challenge each other’s suggestions.
Whenever I use AI to generate an image, I label it.
I use AI carefully for research. When I was researching how data centers might affect electricity bills, I used two AI platforms to generate research reports, then verified their claims against the original sources they cited.
Coming up next
Part 2: How AI guided me to get great care from medical professionals, and how it saved me when they dropped the ball.
Part 3: How I used AI to understand a local energy issue
I’d love to hear from you
How do you use AI? Why do you avoid it? How has it been helpful? How has it disappointed you, or worse?
Reply with your own AI story. One sentence, one experience. Leave a comment or reach out to me here.
See the latest Northfield.org posts in the archive.
I have just begun to use AI. Two tasks, suggest an index for my memoir which was helpful and encouraging and modify some photos by replacing faces which didn’t work very well and changing a photograph to a painting which did. Mostly, I see a need to develop my instructional skills. Having fun learning.
When I think “AI”, I see Arnold Schwarzenegger kneeling in an alley as the cyborg sent by AI of the future “Skynet” to ensure the extinction of the humans who created it. 😱