How I'm Using AI

AI isn't perfect, but it's ready for prime time—my hands-on examples from podcasts and theology to home repairs, and why non-users risk getting left behind.

Following up on my podcast with Dean Ball on AI, I wanted to share a little bit about how I’ve been using it.

My early forays into trying out AI chatbots were disappointing to say the least. Their training data stopped more than a year into the past and they couldn’t access the web, so they were useless on current events. They regularly made stuff up. They couldn’t supply links to sources that substantiated their claims (at least not real links at any rate).

About a year ago I decided to adopt a policy of “AI first.” Meaning, I wanted to try to use AI as a solution before using more traditional approaches. In practice, this meant using AI (usually Grok) rather than Google for information searches. It’s amazing how instinctive firing up a Google window is. It took me months to just break the almost autonomous habit of just searching Google. And it wasn’t necessarily any easier either. With Google, I’d just use keywords to try to find things, whereas with AI I was typing in questions. Especially on a phone this was more painful.

But in recent months, AI has become much, much better than it used to be. Though while not without errors - it continues to make things up, aka “hallucinating” - it’s already super-useful and is only going to become more so. I have subscriptions to both Claude and Grok, and regularly use both.

I typically use Grok to spell and grammar check my posts before publishing them. It also gives me headline suggestions. For podcasts, Substack’s built in AI transcribes the audio of my podcasts. I can then feed the transcript into Grok, which will create Youtube descriptions with suggested chapterization and timestamps in the correct format. I use an AI product called Eleven Labs to generate the audio versions of my posts. I had Grok generate the cover image for this post.

I also use Claude a lot to help with research or developing ideas. It’s very helpful in finding examples to illustrate points I’m trying to make. I also use it to help refine ideas and frameworks. It will give me feedback and suggestions about things I might be missing. It can also fact check things for me and flag potential problems.

In short, I don’t use AI for any actual writing, but employing it extensively in the tasks that surround writing.

These chat bots are also very good search type engines these days. They can provide links to real time articles. I only really use Google these days when there’s a very specific web page I’m looking for. For example, I still use Google to quickly get links to my own articles I’ve written. I can search by title on this site to get them.

Claude and Grok are also fantastic for getting backgrounders on topics. Think of them as an interactive encyclopedia. I recently read my first book by René Girard, and thought his interpretation of the death of Christ was pretty heterodox, and probably more aligned with Eastern Orthodoxy than Western Christianity, despite Girard being Catholic. Grok gave me all kinds of information about how various people had responded to Girard, as well as a comparison between his views and Eastern Orthodoxy.

AI is great for getting a summary of how various commentators have interpreted passages of scripture. One of my favorites things to do with it is, when I hear something from a pastor that sounds like it might be a modern innovation, to ask whether or not any major figure in Christian history taught this prior to 1900. Out comes an entire history of the idea - if there is one.

If I want to use any of this material in print, I have it spit out links to credible sources.

Musk has integrated Grok right into Twitter/X. You can click the Grok icon on any tweet and get an explanation of it, and the context of the post. (I think it’s actually regressed a bit here, but if it doesn’t explain everything, you can prompt it for exactly what you are looking for). If someone simply posts a screen shot of a paragraph out of an article or book, Grok will often tell you the source and even provide a link to it for you.

AI has also proven useful for tons of personal tasks like home repairs. I can post a picture of something and have it diagnose what’s wrong and provide instructions for how to fix. In some cases this can be way better than the old search methods, because it’s interactive. You can tell it what you’ve tried and what you are doing, and it will guide you along or make suggestions. It will also tell you if an option you plan to try is a bad idea or not. Even if it misses some, it will at least catch a few.

I’ve also been using Claude Code, which is a command line “vibe coding” tool. Claude Code is like having superpowers. I read in Neiman Lab that the New York Times had built a tool they called the “Manosphere Report,” that produces transcripts of manosphere podcasts. I had Claude Code whip one up for me. It downloads any podcasts I tell it to (from Youtube, Rumble, or Apple Podcasts), transcribes them, then produces a summary report with guests, topics discussed, and any controversies. The whole thing is automated. It took me two hours to create with Claude Code, 75% of which was me trying to figure out how to install all the dependencies.

I’m using it for other things as well. This is so powerful that I haven’t even figured out all the things I can do with it. I think I’m suffering from a lack of imagination of all the ways that AI will allow me to do things that aren’t even in my mental toolbox right now.

My next goal is to get to the point where I’m using AI so much that I’m fully using a Claude Max subscription. I figure if I’m not using it well beyond the basic account limits, I’m not using it nearly enough.

In short, AI is real, useful, and ready for primetime in some applications right now. You do need to validate the output and other things. It’s not fire and forget. But there’s a ton it’s already doing.

Just the improvements that AI has experienced in the last year are incredible. There’s every reason to believe it’s only going to keep getting better. That’s not saying there won’t be a big AI financial crash, lots of failed companies, etc. But this technology is very much for real.

If you aren’t using AI yet, I really suggest leaning in hard to trying it and figuring out what it can do. It’s a technology you don’t want to find yourself behind the curve on.

You might also be interested in this perspective on AI from Ken Corless. He’s a former Accenture colleague of mine, now at Deloitte, who has been leading complex technology implementations for a long time. I believe he just oversaw the technology implementation at the Milan-Cortina Olympics. He wrote an interesting piece “Ten Things I Think I Think About... AI, Agents, Humans and Software Development,” which is an interesting perspective from inside the IT consulting industry.