Be honest about how much writing actually takes

We complain about AI slop, then build a publishing economy that makes doing everything manually increasingly irrational.

There is something slightly strange about the conversation around AI-generated writing.

We complain about AI slop.

We celebrate things that feel handcrafted.

Then we tell people they should post three or four times a week on LinkedIn, publish a newsletter, maintain a website, comment on other people’s posts, build a personal brand and somehow do their actual job as well.

Be honest.

How is that supposed to work?

I’ve been writing professionally for about 15 years. I know how long writing takes.

Not typing. Writing.

Thinking of the idea. Working out whether there’s actually an argument. Finding the opening. Structuring it. Writing a bad version. Rewriting it. Cutting things. Checking things. Reading it again the next morning and wondering why you wrote it like that.

I’ve done plenty of it manually.

That’s partly why I’m comfortable saying I don’t particularly want to do all of it manually anymore.

I wrote recently that AI writes a lot of my posts. This is the economic argument underneath that workflow.

If I’m working full time, have a family and want to develop Editorial Intelligence alongside that, there is no realistic version of this where I spend two hours lovingly crafting every LinkedIn post from a blank document.

And I’m increasingly suspicious of the gap between what we say about AI-assisted writing and what the economics of publishing actually demand.

The publishing economy rewards frequency

Look at LinkedIn.

Most posts disappear quickly. If you want your ideas to surface, you need to publish enough of them for people to encounter you repeatedly.

One brilliant post every six weeks might be beautifully handcrafted.

It is also quite likely that hardly anyone will know what you think.

The same problem exists with newsletters and platforms like Substack. Some people publish substantial pieces remarkably frequently.

Some are full-time creators. Some have teams. Some have developed very efficient workflows. Some will be writing everything manually.

And increasingly, some will be using AI.

That isn’t an accusation. It seems like an entirely rational response to the production economics.

The useful question isn’t whether every prolific writer secretly uses a model. We don’t know, and plenty of people can write very fast.

The stronger point is that the economics increasingly favour assisted production whether people feel comfortable admitting that or not.

The more channels we expect one person to maintain, and the more frequently we expect them to appear, the harder it becomes to insist that every sentence should begin with a blank page and a blinking cursor.

”AI slop” can become a comfort button

This is where I think the phrase “AI slop” becomes slightly too convenient.

There is plenty of AI slop. Generic writing has become extremely cheap to produce, and the internet is going to contain much more of it.

But humans were producing forgettable corporate content, SEO filler, generic thought leadership and formulaic LinkedIn posts long before generative AI arrived.

AI has made producing that material dramatically cheaper.

It didn’t invent the underlying problem.

Sometimes calling something “AI slop” seems to let us resolve the whole argument at the level of prose. Spot a familiar rhythm. Notice an em dash. Decide the machine was involved. Press the imaginary slop button and move on.

It makes the judgement easy.

But it avoids the more useful questions.

Where did the idea come from?

What evidence sits behind it?

Does the person publishing it actually believe it?

Did anything happen in the real world that caused them to think it?

Can they defend it when somebody disagrees?

Has the machine replaced the thinking, or has it reduced the production cost of expressing the thinking?

Those are very different things.

Not everyone was trying to become a writer

A lot of the loudest complaints about AI writing come from people who write for a living, or who built a professional identity around writing well.

That’s a reasonable thing to be protective of. It took years to build.

But most of the people now expected to publish were never writers. They’re accountants, consultants, engineers, founders, people between jobs. Nobody asked them to get good at prose. They were asked to demonstrate expertise, and publishing became the way you’re expected to do that.

Holding them to a professional writer’s standard, then calling the result “slop” when it falls short, isn’t really a quality argument. It’s a craft applying its own bar to people who never asked to join it.

And there’s no realistic way to enforce that bar anyway. A huge number of people who were never going to write well by hand now have a way to get their thinking into publishable shape. Telling them not to isn’t an option that exists. Nobody with something worth saying is going to sit it out because they can’t turn a sentence the way someone who’s done this for 15 years can.

That doesn’t excuse the actual slop — thinking that isn’t there, publishing with nothing behind it. But the standard has to be about whether there’s something real underneath. Not about whether the prose sounds like it came from a professional.

There is a difference between producing words and producing thought

One reason the conversation gets confused is that we keep collapsing several kinds of work into the verb write.

There is noticing something worth exploring.

There is developing the argument.

There is gathering or retrieving evidence.

There is connecting the idea to previous work.

There is deciding what you actually believe.

There is structuring the piece.

There is producing sentences.

There is editing those sentences.

There is deciding whether the thing deserves to be published under your name.

AI can participate in most of that workflow now.

That doesn’t mean every part has equal value or that handing every part to a model produces good work.

For me, the important line is the one I described in AI writes a lot of my posts: am I acting as the managing editor of my own thinking, or merely approving something plausible that a model generated for me?

I want AI doing a lot of production work.

I don’t want it quietly deciding what I believe.

My workflow is built around that distinction

My own process increasingly looks like this.

I notice something.

I talk it through.

I connect it with other things I’ve written, researched or experienced.

I use the Editorial Intelligence material I’ve been building as context.

AI helps me interrogate it, structure it and draft it.

Then I edit.

Sometimes heavily. Sometimes surprisingly little.

The important part is that I can trace the finished thing backwards.

There was an observation.

There was thinking.

There was context.

There was an argument I was willing to put my name against.

AI made getting from those things to a publishable object considerably faster.

That matters because the work happened before the prompt. The apparent speed of the finished output can hide years of interviews, writing, reading, mistakes, professional experience and previous arguments that made the prompt useful in the first place.

And the more of that accumulated thinking I preserve, the more useful it becomes. That is the argument behind context is capital: AI doesn’t only reduce the cost of producing new work; it can increase the value of work already done by making it easier to retrieve, compare and reuse.

The alternative is often not better writing. It is silence.

This is the bit I think gets missed when experienced writers argue that people should simply write everything themselves.

The alternative isn’t necessarily that I spend every evening sitting at a desk carefully composing essays by hand.

The alternative is that most of these ideas never get published.

That matters when you’re trying to build something while doing another job.

It matters even more when you’re looking for one.

We’re increasingly telling people that their public body of work matters. Show what you know. Build a reputation. Demonstrate expertise. Develop a network. Publish your thinking.

Fine.

But somebody searching for a job also has applications to make, interviews to prepare for and probably a job they’re still doing.

Are we seriously going to insist that they manually manufacture every word of the body of work that demonstrates what they can do?

I don’t think that’s sustainable.

AI changes the production equation.

It means one person can potentially operate more like a small editorial team.

That creates enormous quantities of rubbish.

It also means somebody with years of experience, a body of evidence and something they genuinely want to explore can finally publish at roughly the speed they think.

Those two outcomes come from the same technology.

Quantity still isn’t the goal

There is an obvious objection to all of this.

If AI makes production cheap, why not publish 10 times more?

Because more is not automatically better.

Frequency matters only if there is enough worthwhile thinking to sustain it.

Using AI to turn one weak observation into six posts, a newsletter, a carousel and an article doesn’t create six times the insight. It creates distribution around a weak idea.

The danger is that once production ceases to be the bottleneck, people start treating the ability to generate as a reason to publish.

That is backwards.

The editorial question becomes more important, not less.

What deserves to exist?

What is genuinely new?

What is merely a variation of something already said?

What has enough evidence behind it?

What is useful in this format rather than another one?

What should be left in the notebook?

Cheap production increases the need for judgement because it removes one of the natural brakes on publication.

This is why the words were never the whole value

If your professional value rests entirely on being the person who manually turns ideas into grammatical sentences, AI is uncomfortable technology.

I understand why.

But I increasingly think the words were never the whole value.

The value can sit in recognising the signal, knowing which question matters, understanding the audience, remembering the relevant evidence, seeing the connection, rejecting the easy argument, making the editorial call and taking responsibility for the result.

Those capabilities don’t become less important when sentence production gets cheaper.

They become easier to see.

Which is why I’m less interested in asking whether AI wrote something.

I’m becoming much more interested in asking whether there was anything worth writing in the first place.