AI-augmented Writing from Zettelkasten

Edmund

Writing with my Zettelkasten has taught me patience. Every time I start a new piece — a post, an essay, or even a book — I need to build a new linear order from a network of interconnected ideas. That means sorting through permanent notes, shaping a storyline, and finding the right tone for the audience.

It’s slow work. As William Zinsser said, “Writing is hard work.” But could it be easier? We now live in the age of AI. One simple prompt could produce the desired output.

Let’s imagine two options:

Option 1 – Let AI write for you.

“Please provide a post for the Zettelkasten.de community about ‘AI-augmented Writing from Zettelkasten’ to start a discussion about the use of AI for writing.”

The concept sounds brilliant. It saves time, delivers new insights at almost no cost, and feels incredibly efficient. Yet one question remains: how does my writing improve? The work is done by AI — not me. No ideas from my Zettelkasten. No test of my thinking or style.

Option 2 – Let AI act as an editor.

Here, AI doesn’t replace the writer; it joins the revision cycle. The process is simple:

“Please share your critique for my given draft.”

With every loop, the AI produces feedback on the current draft. I decide what to change, revise, or ignore. When I’m satisfied with the result, I stop the cycle and publish.

This post is my third and final draft — created with AI for editing. Now I invite you to experiment with Option 1 and/or Option 2 using your preferred AI tool.

To close with another line from William Zinsser:

“Ultimately the product that any writer has to sell is not the subject being written about, but who he or she is.”

How do you use AI in your writing workflow — as a writer, an editor, or a silent observer?

References

• Zinsser, William. On Writing Well, 30th Anniversary Edition: An Informal Guide to Writing Nonfiction. Harper Perennial, 2014.
• Opperman, Deon. dir. How Much AI Makes You an AI-Augmented Writer? 2024. 09:17. https://www.youtube.com/watch?v=xW_PxJeEf5c.

andang76

My two cents.
I use AI only for some cognitive works when I'm very tired or I have exhausted my energy doing a thing (for example, I've translated my long post into the other thread from italian to english after hours of thinking and writing, I simply didn’t feel like taking on that commitment too), or when I’m really stuck on a thing that needs a push.
As a general opinion, the more I let an AI do cognitive work for me, the less I train myself to do that specific well. If I overuse it, I’ll lose the skill. It’s a matter of trading time for developing or maintaining a skill. I'm aware that every piece of processing an AI does is a piece of processing I’m taking from someone (something) outside myself. So, only if strictly necessary, not as a "better way to do things" .

KillerWhale

I f*cking don't and I want to see those things die in a fire. I was at PAX Australia this week-end and as a music composer on a panel eloquently put it: "If you use Sora, you're not a composer, get the hell out of here."

AI is based on predatory practices, messes up the environment and kills my own capacity for improving my own thinking through the sheer process of doing stuff. Every time I've tested LLMs I came out with a sour taste in my mouth and the feeling that a few of my neurons had died.

The end goal of writing is not in having written, it's in the process and what you learn out of it. Wanting to write with AI is like having a bot going to the gym, or to church, in your stead and bringing you back what you should have taken out of it (and messing up the message in the process).

It's. Not. The. Point.

zettelsan

@KillerWhale said:
Every time I've tested LLMs I came out with a sour taste in my mouth and the feeling that a few of my neurons had died.

LOL, and I do share the sentiment. LLMs do make you more efficient in many ways but don't make you more capable.

And that's why the relationship between AI and humans will be a major conundrum.

If it's about self actualization, AI will take that away from the individual, so he may choose not to use it. It's an individual choice.

If it's about productivity in terms of outcomes and results, however, the world may not care who produces them. If AI can do it more efficiently and cheaply, whoever benefit from it will choose AI over a less efficient person who seeks self actualization.

I do worry how that plays out going forward.

Sascha

@KillerWhale said:
The end goal of writing is not in having written, it's in the process and what you learn out of it.

This is where I have to disagree partially. For a lot of writers, the goal is to have written a piece, while the internal change is just a necessary evil. Students who are forced by the curriculum to hand in stupid essays. Students who need to study because of certificates that don't prepare them for the actual job well (in education). Companies (who formerly hired copywriters to produce the slop) want to do "content marketing". Many who want to build a social media presence.

There are a lot of use cases in which you just want the text to be written in any way.

The comparison with the bot obviously holds.

I think it is simply not an efficient way to actually produce something that builds beyond known knowledge. In my case, this leads to:

• AI is often not able to translate my work properly, since I have a particular way in German to use words, which seems to be at odds with the predictive process underlying LLMs.
• The only cases in which I could actually let AI write something are the first draft of the summary of articles, with AI trying to be as faithful to my writing style as possible (via prompts and source material). But then I'd miss the thinking effort of summarising, and I still need a lot of training in writing.

But for creative work for fantasy world-building, AI is strangely productive. But it rather provides me with material that I can process, and not any results that I can use.

But I don't think the world is a worse place if "spicy" romance novels were taken over by AI, with a few genius-level works peaking through the mass of slob. (Which is already happening, btw.)

The process of automation has already taken place in a lot of the physical realm, with machines having taken over work like farming (which replaced hunting and gathering) and manufacturing. Now, we train and diet to make up for it.

So, I say let AI take care of the slob that is already produced and occupies people who could do more productive things.

I agree with the mechanics, but not with the existential foundation. I think AI doesn't work as intended in many cases.

KillerWhale

@zettelsan said:

LLMs do make you more efficient in many ways but don't make you more capable.

Thank you for that, I think you've absolutely nailed the core of the matter. The philosophical issues and the social implications probably all boil down to this.

I write because I do want to produce an end result, but I've always been on a quest to increase my abilities and understanding – this has been an incredible side effect of being a writer for decades at this point, and I've found it to be a crucial important aspect of it. Which is why I despise LLMs so much.

Writing is like playing an instrument. You play it because it makes sense to you, and maybe people will come listen. If you have to play it because it's a necessary evil, then something is already lost. And of course, it does happen – but a) having people being forced in that situation and b) lacking the proper tools and training to properly deal, is one of our many failings as societies.