Last August I quietly scrubbed the em-dashes out of a post. A year on, I use AI more than ever, and I've stopped worrying about it.

In my commit history of this blog last year, there was a small commit that I titled De-AI ... markers. All it did was change two sentences in an earlier post. It replaced a couple of em-dashes with parentheses and commas, and rewrote my words from “not only X, but Y” into something plainer. Nothing else in the post changed.

At that time, it bothered me that my writing wasn’t entirely mine anymore, that somehow AI had put words on the page that weren’t solely my own. I didn’t think AI had written any of that piece, but it had certainly become my editor. The em-dashes were the tell. And the tidy sentences. I remember scanning through and feeling a little uneasy about who was actually doing the writing.

A year on, I use AI even more than I did then. The honest answer is I’m more enthused about relying on it now than I’ve ever been.

At work, when I’m writing code by hand, I easily make typographical and reasoning mistakes. Refactoring across an unwieldy codebase is another kind of tedious work that makes no sense to do manually. For both, the AI just does it, faster and more accurately than I ever could, and I get to spend my time on the parts I do enjoy: reasoning about the outcome, and keeping the structure clean. Things that used to take me hours or days now take a fraction of that.

At home, I use it for Spanish learning. AI is a much more interactive tutor than any textbook I’ve read. It’ll explain why a subjunctive shows up somewhere, quiz me on it, and let me ask embarrassing follow-up questions without judgement. (Although for listening practice, I still rely on podcasts.)

And for this blog, it fixes my grammar (not to mention that it patched a broken pagination on the homepage last week). Fixing that manually would have taken hours of frustration. Today, I care less if a chatbot helped fix my pagination or smoothed out a sentence. The intent, design and message I want to convey are still entirely mine.

So what’s changed is that I’ve realised there will never be a day when I can express myself better without a tool-editor to fix up my typos and grammatical mistakes, and having one lets me share my thoughts more easily than ever, without worrying about making mistakes. The AI can phrase almost anything better than I can, but it doesn’t know what I’m thinking about this week, or what my Fitbit has been telling me about my sleep, or how I actually felt about turning 30 with $150k in the bank. The AI can lend me the words, but the thoughts are entirely mine.

So the deal I’ve quietly made with myself is this. I write the first draft, and I keep the clumsy ideas in, because they’re the proof that I was the one doing the thinking. Then AI comes in and tidies up my grammar and typos, and even rephrases where it helps. At the end of the day, these are still my personal thoughts. And you’ll be thankful too that you don’t have to wade through spelling errors or clunky sentences just to get to what I’m actually saying.