Confessions of an AI Addict

I never intended to be an addict. Nobody does, I suppose. But the temptations, they’re too powerful. They’ve made me lazy, dishonest, selfish, needy, over-confident — every sin under the sun.

I’ve tried to resist. Many times. And sometimes I escape, for a little while. But it always pulls me back in, usually sooner rather than later.

My addiction is to algorithms, to AI. It’s my digital cocaine. It makes me feel invincible while actually numbing my brain and distorting reality.

Like every addiction, this one started innocently.

I played around with DALLE and ChatGPT over Christmas in 2022. Soon, I was asking it questions about work. Within weeks, I was letting AI write drafts of headlines, intros, and conclusions — the path to my first little sin.

ChatGPT came up with a good hook and a great joke. I slipped both into a draft. When the piece came back from my editor, the intro had a compliment and the joke a laughing emoji.

I took the credit, and a pattern of increasing, undisclosed AI involvement took hold.

It also marked the beginning of my first period of over-confidence in AI’s capabilities (and it wouldn’t be the last).

I spent hours, sometimes days, in algorithmic alchemy, tinkering with workflows I hoped would deliver articles at the click of a button.

They rarely did, back in 2023. I’d find myself scrambling to get my work done in time after yet another grandiose AI scheme let me down.

After weeks or months of this — I don’t remember exactly — I entered my first period of AI abandonment.

Disappointed with the results, I started using ChatGPT only as an infrequent sparring partner. I mostly returned to my old ways: intense reading and highlighting, heavy-handed organizing of research, obsessive writing and rewriting.

But on March 3rd, 2024 (I remember exactly where I was; see picture), I discovered Anthropic’s Claude and its Opus model. Unlike ChatGPT, it could handle lots of information, seemed a better writer, and just generally felt more capable.

I restarted my alchemic experiments, and now they seemed to work.

I knew, because, at the time, I committed probably my biggest sin: I ghostwrote someone’s guest posts and the author signed off — without feedback and with compliments. Unbeknownst to them, Claude was the real ghostwriter and penned 90% of the work.

That “success” triggered a new wave of AI usage and, ultimately, dependency. It wasn’t until working on Stay Strong: Never Let AI Fill Your Blank Page that I realized how much I’d been pulled back into the algorithm’s grip:

I was relying on AI to do a lot of my “reading.”

I let it do lots of outlining. Ideation. Drafting. Addressing feedback.

My brain got lazy. The original thinking, the hard work… it was all gone.

I don’t have much to show for that period (except, ironically, those “guest” posts Claude wrote for me).

Stay Strong forced me to pause and reflect on my behavior.

I decided to revert to my old work habits for that piece. I did the research, sweated every word, and then still rewrote sentences and killed darlings until right before hitting Publish.

The article took forever to complete, but it’s also one of the best things I’ve ever written.

The main thesis is that you, the human, should always fill “the blank page,” before involving AI. Otherwise, if you let the algorithms go first, they’ll anchor your thinking to their ideas, or worse, lull your brain to sleep.

“The blank page” is literal for writers, but it’s a metaphor for other domains, too.

In any endeavor, you need to do the hard thinking first. Don’t shortcut through that struggle with AI.

Writing that article firmly planted this mantra in my head. Still, I often fail to follow it.

That’s the strength of AI’s pull, and it’s only getting stronger. All the alchemy I imagined doing with AI back in 2023 is now possible — and much more:

OpenAI’s o3 model is more intelligent than I am in many areas.

GPT-4.1 and Claude Sonnet are in the 80-90% range of mimicking writing styles correctly.

Cursor lets me build things I never would have been able to before (like SEO traffic calculators).

Deep Research provides research reports in minutes that would have taken hours or days without it.

As these capabilities expand further, so does the dependency, the addiction, and my stupidity.

Just recently, I’ve hosted podcast interviews without much more preparation than reading the AI-generated questions. I’ve sent piles of information into AI workflows without reviewing them. I’ve clicked Ok on algorithmically created outlines without adding an original thought. And I’m asking for feedback on everything, all the time (the neediness!):

Is this email ok?

Can I say this?

What do you think of this?

How did I do in this meeting?

Life without algorithms will become impossible for most of us, even if we hate to admit it.

Like smartphones and social media before it, AI will pull us all into its clutches, but in an even more dramatic and fundamental way. It will penetrate our brains and change how we think, behave, and live.*

I’ve seen coworkers dump verbatim AI output on others more than once.

I’ve heard people blame Claude for mistakes in their work.

I’m seeing AI fingerprints all over my LinkedIn feed.

Such a force can only be tamed with serious counter measures — like the ones I took to write this article: no AI allowed during any part of the process, and hand-writing the first draft.**

I wanted to feel how dependent I’d become. I wanted to know if I could still write, and think. Most importantly, I wanted to observe my own addicted mind, to see what it would do without AI.

Going cold turkey quickly exposed several hiccups in my AI-saturated work process.

I wanted to escape the wrestle of finding perfect words (“I’ll discuss that with ChatGPT later”).

I was too lazy to create good titles, MECE lists, and airtight arguments.

I kept wanting to ask AI for feedback on everything I thought and wrote.

I had to find my way back to Google’s blue links to find information. (I’d gotten spoiled with ready-made answers from AI overviews and Perplexity.)

I had to plan ahead to involve a designer for the images**, and leave enough time for our editors to process the piece.

To my surprise, it took less than an hour to get used to these “inconveniences,” and the benefits of my Luddite approach became clear just as fast.

The experience reminded me of how, in Hamlet’s BlackBerry, author William Powers describes the effect of losing his phone on a boating trip:

“There I was with no inbox to check, nothing to click on or respond to. No demands, requests, or options. No headlines to scan or orders to place. No crowd to keep me busy. With all of that out of reach, my consciousness had no choice but to settle down into the physical place where I happened to be and make the best of it.”

Especially that last part. With no AI to consult, my mind settled down. After the initial withdrawal symptoms, I realized I could still write! My brain could still think!

I reached depths of thought and states of flow I hadn’t experienced for a while. And every time I did a word count after a half hour hand-writing session, I was pleasantly surprised: 4-5 pages, 500-600 words total.

But I don’t want to ban AI from my life.

It does give me amazing capabilities. I wouldn’t want to lose vibe coding, deep research, and having a 24/7 sparring partner (in moderation). But AI comes with a dark side that corrupts the mind, like The One Ring, cocaine, or the Sirens.

Unless you want to choose between all or nothing, you’ll have to be like Ulysses and tie yourself to the mast. You need to acknowledge you’re playing with fire stronger than your willpower. Something that, like a hard drug, will mess with your mind and trap you into its spell if left unchecked.

And so these are my vows, the steps I’ve started following, to try and give my mind a chance to tame — or at least co-exist with — the beast.

1. Write something by hand, daily

There’s no blanker page than the one you have to fill by hand. It feels tedious, but it’s a highway to the depths of your mind.

Writing by hand ≠ writing articles. You can handwrite your first thoughts about a strategy, a new campaign, or a difficult situation. The point is to let your mind think through a pen at least once a day.

2. No AI dumping on others

AI dumping turns your laziness into someone else’s problem. Because you’re neither thinking nor curating an AI’s output, you overload someone else with slop to wade through.

If all you do is dump the output of a prompt on someone, better send them the prompt (so they can understand and refine the input) or do nothing at all. My new mantra: “I’d rather get your prompt than your dump.”

3. Break free regularly

Define moments when you’ll work without AI. First hour of the day, after lunch, evenings — whatever works.

Like a digital fast or detox, you need moments in your day and week where you can disconnect from AI’s power and speed. Moments where your mind can settle down. Schedule such breaks intentionally, or they’ll never happen.

4. Know when to pay attention, then do

Every process has critical points where human thinking really matters: the thesis of your article. The big idea for a customer pitch. The point of view that underpins your thought leadership.

In such moments, AI pulls its biggest tricks, convincing you it did a great job, seducing you into laziness, and letting algorithms take over the wheel.

You can not let that happen. These are the junctions in your work that need screaming warning signs for your attention, energy, and creativity.

5. Read the research

My pre-AI self would never believe I would have to write this rule, but it’s true: you can’t skip reading the research. Not just to catch hallucinations, but to understand the topic, have a point of view, and be able to judge the AI’s work.

Unbelievably, I’ve committed this sin too many times under the influence of AI. Perhaps it’s the hallmark of a true addict. So read the research — and AI summaries don’t cut it, of course.

6. Call out AI sins in others (and let yourself be called out)

AI addiction isn’t only about your behavior or mine. Part of the problem is our collective complicity in this AI-driven descent into mediocrity.

We see a coworker AI dumping on others and say nothing.

We know someone’s LinkedIn post is a half-formed shower thought propped up by an algorithm — and still we hit Like.

If we accept mindless AI garbage everywhere, it becomes the norm.

Like a group of friends where everyone’s doing cocaine all the time, reality gets distorted, and addiction doesn’t seem like addiction anymore.

So it is with AI. Even when you’re able to stay clean — with or without serious counter measures — you must also hold others accountable. Call out sins when you see them, and encourage people to keep doing the hard work that creates originality, meaning, and value.

Only by supporting each other can we wield these great new powers and reach an equilibrium between abstinence and addiction. So please, hold me accountable when I lapse again, and I’ll happily do the same for you. ❤️

_{* In the personal realm, I’ve already relied on AI for marriage advice (see this article in MIT that features a certain Tim).}

_{** I did write on an e-ink tablet, but no AI was involved.}

_{*** I didn’t plan ahead far enough — our designer wasn’t available in time, so I had to come up with some visuals myself. 😀}