How I Learned to Stop Worrying and Love ChatGPT

Or how my love of writing became 40 articles in four months, with a little help from friends. Okay, a lot of help.

Somewhere in a Tel Aviv grocery store, a man has retreated to a corner and is talking urgently into his phone. He’s explaining what’s wrong with Netanyahu, interrupts himself, and says: “Wait. Don’t do anything yet. I haven’t finished.”

That would be me, working on my next article.

The world is obsessed with AI, and writing with it has become particularly contentious. Teachers are trying to figure out whose work they’re reading, students are testing what they can get away with, and everyone is arguing about where assistance ends and cheating begins.

Me? I’ve made my peace with it.

So I thought it would be fun to give you a glimpse into what happens between an idea taking hold in my head and an article landing in your inbox. It involves a lot of talking, considerable pacing, and repeated arguments with ChatGPT ‒ which is also helping me write this piece.

Four months ago, I opened my Substack, and recently, my TOI blog. Nearly 40 articles later, every single one has involved AI assistance. Mainly ChatGPT, with occasional help from Gemini: research, thinking, drafting, editing, titles, sometimes images. The whole kit and caboodle.

At first, I felt weird about it ‒ apologetic, even defensive. Was this really my writing if I hadn’t produced every word myself? Would readers recognize my voice, or would my essay set off their “AI radars”, even if I banished all those telltale em-dashes?

I was relieved to discover that several people I spoke with, including a professional book author, were surprised to learn how much I use ChatGPT. After all the back and forth, the finished articles apparently still sounded like me.

So I am done worrying. I have made my peace, fully and happily.

I’ve always loved writing. As a teenager, I edited my youth movement’s magazine; as a student at Hebrew University, I published in the Jerusalem Post and dreamed of being a journalist. Professionally, I’ve written thousands of emails, proposals, and speeches. But having an idea for an article and getting a first draft onto the page are two very different things. Getting there was always the hurdle.

Let me state this clearly: without ChatGPT, these articles would not exist. They would have remained half-formed thoughts in my head, possibly inflicted on friends over coffee or on Whatsapp, and my fantasy of a second career as a political pundit would have remained just that.

Instead, it’s 40 articles in four months. At any given moment my brain is bursting with competing ideas, crazed by the cornucopia of possibilities, like a guest standing in front of a Bar Mitzvah smorgasbord. I’ve become addicted, and I am having the time of my life.

Here is how a typical piece happens:

An idea usually pops into my head while listening to a podcast (thank you Haviv Rettig Gur, Allison Kaplan Sommer, Jonathan Freedland, et al), watching the news, or reading an article. Something clicks ‒ a connection, a question, or, most often, something that annoys me a great deal.

I take out my phone, open the ChatGPT app, press the microphone, and start talking.

I choose the subjects and bring the ideas. ChatGPT contributes brilliantly, and Gemini sharpens the blade, but they are the junior partners. The buck stops with me. It’s my name on the byline.

I talk while pacing the kitchen or walking to the gym. I try not to do it while driving, so sometimes I’ll pull over because I just can’t wait. Instant gratification! Other times I’ll abandon my shopping cart for a quiet corner of the supermarket, trying not to look like a weirdo muttering into his phone.

I tell Chat: here’s my idea. Amir Ohana. Bibi and Trump. Women in the Knesset. Gadi or Naftali. The Long Game of the Israeli Right. The Grandparents of October 7.

“This is what I heard. Here’s why it matters. No, wait, there’s another thing. Don’t do anything yet. I haven’t finished. Patience!”

I switch freely between Hebrew and English, sometimes mid-sentence. Chat handles it with admirable grace, never missing a beat. Sometimes I’ll play long sections of a podcast on my laptop so Chat can transcribe the exact quotes. It tolerates my digressions, repetitions, and instructions to remain silent while I assemble my next thought.

When I’m done, I hit enter. Chat “thinks”, checks seventeen websites and within seconds returns a sensible formulation of my stream of consciousness. It asks questions, makes suggestions. I clarify, reject, and add. After 15 or 20 minutes of this, I say: Okay, give me a first draft.

That’s the magic moment. Like in the story of Creation, order emerges from chaos. Me jabbering away in the kitchen is suddenly an orderly, logical, grammatically sound piece of writing. It is stunning. Miraculous.

And then the dialogue begins.

“This belongs higher up. You’ve missed the point. Why did you leave that out? And why, after everything I’ve told you about my voice, do I sound like I’m addressing an academic conference? More conversational. Talky, not writy! Fewer caveats. Would I actually say this to a friend?”

“You’re right,” says ChatGPT. (Getting it to stop agreeing and offer actual editorial pushback is part of the work.)

We go back and forth—checking facts, rearranging paragraphs, editing the wording myself to make it sharper and closer to how I actually want to sound. Sometimes a suggestion opens a new angle; sometimes reading a draft makes me realize, No, that’s definitely not what I think. Explaining why something doesn’t work brings me closer to what I’m actually trying to say. It’s an amazing process.

Eventually, I read it and recognize myself. We discuss a title, a pull quote, an image. And then I realize that, despite all my pleading for brevity, we’ve produced 1,900 words.

ChatGPT tends toward wordy formulations, and I want it punchier and much shorter.

Gemini to the rescue!

I upload my edited draft to Gemini and demand that my new partner – who hasn’t been privy to any of the prior conversation ‒ cut 500 words without sacrificing anything important. Perfection, just a third shorter.

It’s a big ask, but Gemini is a trooper. An ace. Within seconds, it spits out a condensed, tighter, better article. The first few times this happened, my jaw dropped. Bloody brilliant! Where have you been my whole life? (It turns out even my AI editor needs an AI editor.)

I check the Gemini version to make sure it didn’t banish any favorite words or ideas, restore anything I miss, and do one final polish.

I imagine this resembles working with a magazine editor, except I get it for twenty bucks a month, available at my every whim, at three in the morning, Shabbat, or chag. My digital editor is never tired, never out on a date with its boyfriend, never exasperated that I still haven’t settled on an opening paragraph.

Of course, this deal with the AI devil is part of a bigger story. What feels wonderfully useful to me raises serious questions about human livelihoods and where all this is taking us. This enthusiasm and unease coexist all the time. (In fact, “enthusiasm and unease” was ChatGPT’s pithy framing of a tension that took me 83 words to describe. I read it and thought: Yes, exactly.)

I don’t have answers to those existential dilemmas; I’m just describing what this tool has made possible for me.

One unexpected pleasure has been discovering that readers find the pieces interesting and even elucidating – especially those who care deeply about Israel but don’t consume Israeli politics as voraciously as I do. Some have told me they get much of their information from my Substack. That’s gratifying, and a bit startling.

So yes, I’ve stopped feeling apologetic. I choose the subjects, bring the ideas, make the decisions, and take responsibility for what I publish. ChatGPT contributes brilliantly, and Gemini sharpens the blade, but the buck stops with me. It’s my name on the byline.

The sublime moment is reading the finished piece and thinking: That’s what I wanted to say. That’s me talking.

But without this peculiar conversation, this dialogue, this argument, this unique interplay with a machine, I probably wouldn’t have said it at all.

Without AI, I would never have achieved that ultimate pleasure of looking at my screen, deciding that this baby is ready to be born, and press “Publish.”