🤖This article is part of the Writing with Robots series, wherein I dictate (that is, rant) into ChatGPT, which transcribes and rewrites my words into semi-readable essays.


Welcome to Writing with Robots, a new experiment I’m trying out.

For years, I’ve had ideas I wanted to share—concepts I’ve developed, things I’ve noticed, strange little theories. But writing is slow. Writing well is even slower. Unfortunately, I have little spare time.

Two years ago, things changed. Large-language models (LLMs) became mainstream. They're remarkably good at, er, language. From day one, LLMs have been substantially better writers than most people. This has gotten to the regrettable point that well-written text is now often considered a red flag: "This is too well written, it's probably AI".

One day, I discovered that I could chuck a 10-minute voice note in, and get back a half-decent written article. The lights went on - I can finally publish my ideas without dozens of hours of writing!

And that's how we got here.

I've found that long-form dictation has many additional benefits. It helps me clearly think and organize my ideas. I then have the bot ruthlessly critique them— sharpening, challenging assumptions. It’s not just a transcriptionist; it’s a surprisingly decent editor!

The content is mine. The ideas are mine. But the style is sadly more collaborative. I plead with it to preserve my voice, but it’s no court stenographer. It takes liberties, rearranges my thoughts, and attempt to sneak in its own ideas. The perfectionist in me objects—the pragmatist has deadlines. And most importantly, this process gets the ideas out of my head and onto the page. 

While I do a manual editing pass, the final text is about 60% AI-written. I can focus on the ideas, which is what I actually care about.

And that’s the goal here. I don’t aspire to be a great writer. Nor impress frizzy-haired English professors. I just want to share weird and wonderful ideas – I'd rather think than write!

I'm not interested in convincing people of anything either. In fact, I’m anti-persuasion. Persuasion seldom leads to truth. Charismatic cult leaders and clever advertising firms routinely convince people of all sorts of bizarre stuff, but that doesn’t make what they’re saying true. Quite the opposite–It just means they’re good at convincing people. So I'll try to avoid that, and simply lay out my thoughts for your inspection. Does it make sense? You decide. 

The posts in this series will be eclectic, eccentric, and likely of interest to no coherent audience. They’ll reflect whatever I happen to be reading or thinking about at the time. Some ideas will hopefully be solid. Others will be underbaked, if not downright raw (Ooh, cookie dough!). Either way, I'd like to get some ideas out and see what people think about them.

So: welcome to Writing with Robots. Let’s see where this goes.