Writing = thinking; Oracle fires you by email; fake AI thought leaders: Newsletter 8 April 2026

Newsletter 146. Writing trains minds, while outsourcing it to AI leaves you stupefied. Plus AI as writing collaborator, editors who upload your manuscript to AI, three people to follow, three books to read, and how you can establish that you’re not just data.

Writing is thinking. AI can’t do that for you.

“I didn’t really know what I wanted to say until I tried to write it.”

I hear this all the time. Writing is hard. It forces you to grapple with the value, structure, credibility, originality, and persuasiveness of what you trying to communicate.

Now that you have access to AI, you can delegate lots of your writing tasks to a machine. The machine doesn’t really think, though, and once you’ve outsourced it, neither do you. A failure to think things through leads to valueless, bland, untested ideas. Readers have no use for writing like that.

Here are all the writing tasks you could outsource to AI, and what you’d lose if you did.

Ideation

Do you really have an idea? Try writing it down in a sentence or two. For example, you might write this:

Anyone can be a great manager.

But is that really an original idea? Is it supported by facts? Is it precise enough to be useful to readers? Once you write it down, what may have seemed fascinating when it was floating around in your head looks naked, obvious, or wrong.

Sharpening ideas is hard. Of course, if you ask for help from AI, you’ll be able to brainstorm it. But AI is likely to lead you in a path that either predictable (what everyone else has said) or peculiar (which is where it will take you if you ask it to be “original”).

You’re better off talking to other people and then trying to write down something that’s not a cliché. That will be hard. You might write and delete a dozen sentences until you get one that seems promising.

If you ask AI to do that work for you, you’ll never get to that original idea.

Research

AI is great at research. It’s an extension of what you used to do, which was web searches and literature searches. Not only that, it can summarize what it finds so you don’t even have to read all of it. What have you got to lose?

Wrestling intellectually with existing ideas, that’s what.

You might find that somebody else has had a similar idea to yours. What did they write? What parts of that do you believe? What makes your perspective different? You’re not going to figure that out unless you actually read what the other person wrote.

You may find that there’s evidence supporting your thesis. AI will find you numbers and statistics and save you from reading long research papers. But is that evidence actually credible, or are you citing a study that’s weak and flawed? Are the numbers you want to cite recycled through so many successive references that they obscure the original source’s lack of credibility?

And what about the evidence against your thesis? You need to read those papers to address why they’re wrong (or maybe, why you’re wrong).

Well-done research is hard. Using AI to help with research saves time. Using AI to replace research deprives you of the opportunity to build a solid, vetted framework around your idea.

Structure

A good piece of writing is always a narrative. You’re not just talking about an idea, you’re telling a story.

Figuring out the elements and sequence of a piece of writing is hard. It requires a balance between putting the most interesting content up front to grab attention, and putting it later on to maximize drama. It typically requires a lot of trail and error moving elements around until they form a persuasive narrative.

Figuring out that sequence is hard, but the results are worth it. Once you’ve done that you really understand what you’re trying to say. You can turn it into a speech. You can turn it into a book. You can repurpose it in multiple formats online.

You could outsource that structural work to an AI. You’ll end up with a perfectly serviceable structure. But is it the best structure? Does it include only the elements that create the most persuasive narrative, the ones that resonate the most with you? You’ll never know if you don’t do that work.

Words

Writing demands that you type words, one after another to form sentences and paragraphs. For lots of people, typing all those words is a really hard. Surely, there’s an ideal way to write any concept, and an AI could do that.

Of course, by depriving you of the opportunity to decide which words to use, the AI removes the most important ingredient in any writing: you.

The objective is not write the best text. It is to write the text that’s the best way to reflect how you would define the idea, tell the story, persuade the reader.

Once you let a machine pick those words, they become generic. With no author, their persuasive power is limited. You fail to connect with the reader.

Without that connection, why bother writing at all?

Hard is good

I’ve told you that ideation is hard, research is hard, structure is hard, words are hard. That’s why people are so attracted to writing with AI — to avoid the hard work.

But that hard work is how you turn a raw idea into a narrative that’s credible, persuasive, and most important, personal. Writing is thinking. If you dodge the hard work, you’re not really thinking. You’re just rearranging existing ideas. There’s enough crap like that in the world, we don’t need any more.

Writing is a pain in the ass. But failing to think through writing will ultimately make you stupid. Short-term gain, long-term disaster. Stay sharp. Keep writing. Because even if it’s harder, it’s worth it.

News for writers and others who think

Joshua Rothman writes in The New Yorker about whether AI might move writing forward and broaden its appeal, rather than killing it (subscriber link). But, as he points out, “The same technologies that expand the creative process also threaten to short-circuit it.”

Are editors at publishing houses using AI to summarize and evaluate manuscripts? Of course they are. I just hope they’re smart enough to use local AI tools that don’t upload content that large language models will then use for training purposes.

Oracle sent 16,000 people an email that started, “After careful consideration of Oracle’s current business needs, we have made the decision to eliminate your role as part of a broader organizational change. As a result, today is your last working day.” The good news: they didn’t bury the lede. But when you’re firing this many people, is efficiency really the top consideration?

As Mark Schaefer points out, using AI to fake your way to thought leadership isn’t going to work. You’ll become known as a ripoff artist, not an original thinker.

Three people to follow

Stacy Ennis, M.A., experienced nonfiction book strategist and coach

Dave Gerhardt , managing an active community for B2B marketers

Gretchen Rubin , happiness expert

Three books to read

City on the Edge: Technology, Politics, and the Fight for the Soul of San Francisco by Jonathan Weber (Atria, 2026). How the internet made San Francisco and then almost killed it.

It’s Still OK to Laugh by Julie Hogan, illustrated by Bob Lipski (Beaver’s Pond Press, 2026). A children’s book that helps grieving kids find some of the joy they’ve lost.

Shut Up and Read: A Memoir from Harriett’s Bookshop by Jeannine Cook (Amistad, 2026). How the author launched a successful bookstore at the start of the 2020 pandemic.

Don’t be data

Sign your name to the “Control Your Persona” manifesto.