A Writer's Guide to AI That Doesn't Suck

As I write this, my wife is downstairs bitching to Claude.

The AI keeps producing perfectly adequate content that's somehow still not quite right.

But she’s a writer, too.

And like that judge who was asked to define porn, a writer knows good writing when they see it.

By way of contrast: a few weeks ago, I helped a stone sculptor write his first AI-generated post for his tombstone business.

We crafted a prompt together, and the AI produced exactly what I've come to expect from first outputs: grammatically flawless, logically sound, and utterly forgettable.

But his eyes lit up. "This is amazing!" he exclaimed, and moved to publish it immediately on his socials.

My heart sank. Thirty years of writing has cursed me with the ability to see what he couldn't - the subtle flaws, the missed opportunities.

This was unseasoned tofu. It would need to be thoroughly marinated before anyone would want to taste it, let alone enjoy it.

Henrik Karlsson (one of the most wonderful essay writers I know, and someone you should follow) often writes about writing.

Here is one thing he says that I think about at least once a week, and it merits your careful consideration:

“Having a method, a process, makes my writing more consistently good. This might sound like a win, but it is a problem. The usefulness and beauty of essays follow a power law, where the outliers create almost all the value—so if I make my essays more consistent, I remove not only the mistakes but also much of the upside."

Henrik Karlsson

This is exactly what we're up against with AI. It excels at consistency - at creating perfectly adequate content. But that same consistency strips away the peculiar edges where brilliance lives.

Breaking Free from the Consistency Trap

One of the beautiful things about AI is how it can help us explore new territories of expression. Like a master chef with an infinite spice rack, we can experiment with innumerable flavours and combinations.

Let me share an approach I often deploy when writing with AI, and how I have learned to avoid those samey outputs that characterises so much of what we see on the internet today.

It all starts with thinking

First, I need to think a little before I start prompting. In fact, if you think a little before you prompt, it puts you ahead of 80% of AI-writers.

I’ve found that if I start to prompt before I’ve had at least 3 good thoughts about what I want to create, I’m usually rather doomed.

Because in order to truly co-create with AI, we need agency. And we can establish that agency by laying the groundwork ourselves.

When your Thoughts are Numbered

Here is how I go about structuring the foundations of my writing (I learned this technique from Malcolm Gladwell, by the way.)

When your thoughts come, write each one down until you have spun out the entire thought. Give it a number: 1.

Then proceed to write the next thought: 2.

You number your thoughts as they come, without worrying about their connection to each other.

Your mind naturally wants to build logical bridges between ideas, but we often get stuck because we can’t immediately find a satisfyingly connector between one thought and the next.

And often the most powerful connections aren't immediately apparent.

When you number your thoughts, you free yourself from the tyranny of needing to making sense. You free yourself from the obligation of immediately being coherent.

That random observation about your grandmother's cookie jar might be the perfect metaphor for explaining neural networks – but you won't know until you let it exist independently first.

The Dance of Human and Machine

Once I’ve collected these numbered thoughts, AI becomes my dancing partner in creating coherence. I feed these seemingly disconnected ideas into the system and watch as it suggests possible connections, alternative arrangements, unexpected parallels.

This is where the magic happens – in the negotiation between human intuition and the machine's pattern-recognition capabilities. The AI might suggest a structure I’d never have considered, while my human judgment decides whether that structure serves my deeper purpose.

The Art of Refinement

The refinement process begins with broad strokes. First, I’ll adjust the overall tone. Is this piece meant to comfort or challenge? To inspire or inform? I’ll let the AI help me explore different emotional territories.

The writers eye comes into play here. I’ll start to notice patterns in how the AI writes – certain phrase structures it favours, particular ways it transitions between ideas. This awareness is crucial because it leads us to our final, most important point.

The Essential Role of Experience

Here's the truth about working with AI that few discuss: the machine will create patterns of sameness that feel unique to novice users. That tombstone business post my sculptor friend loved? To an experienced eye, it carried all the hallmarks of first-generation AI writing.

Developing this discernment takes time. It requires you to write with AI, fail with AI, succeed with AI, and most importantly, to develop an ear for when the machine is truly amplifying your voice versus merely mimicking what it thinks you want to hear.

To be a truly effective AI creator you need to get to a point where you are no longer impressed nor seduced by its output.

Keep asking: the clay may have been rendered into a beautiful form, but does it continue to correspond to my evolving intention?

Excellence in AI-assisted writing doesn't come from better prompts alone – though those help. It comes from the hard-earned ability to recognise when the AI's output serves your true purpose and when it's merely meeting the letter of your request while missing its spirit.

The Importance of Discernment

Discernment becomes your greatest asset. It is not about distinguishing good writing from bad, we are well past this point here – it's about developing an almost physical sensitivity to the weight and texture of words, to the subtle variations in tone that separate the authentic from the artificial.

Imagine being at a wine tasting. The novice can distinguish red from white, sweet from dry. But the sommelier detects layers of complexity: the whisper of oak, the memory of rain on limestone, the suggestion of late summer berries. This is the level of discernment we must develop with AI-generated content.

With experience, you begin to feel when the AI is reaching for easy phrases rather than earning its conclusions. You notice when it's skating on the surface rather than diving deep.

Most importantly, you develop an instinct for when it's serving your genuine intent versus merely fulfilling the mechanical requirements of your prompt.

In the end, creating with AI is like dancing with a partner who knows all the steps but doesn’t feel the music.

That’s your job.

Subscribe to continue exploring the intersection of human creativity and artificial intelligence at 10xbetter.ai