The hardest part of content hasn’t gotten any easier

I’ve spent a lot of my time lately building content systems in Claude Code. The gist being that I provide raw material and a series of Claude Skills find the topics, suggest framing, write drafts and eventually output content optimized for the channel where the piece will be published. This is fun work and I find it to be genuinely interesting and very satisfying (when it works!).

This is an abstraction of the type of work I used to do. I used to write articles and now I create systems to write articles. (Perhaps I’ll write a future article about how these systems work.) I don’t think I could build the systems without the experience of doing it the “old fashioned” way for long enough to develop expertise. And for what it’s worth, I think the output is pretty darn good. 

But I’ve had the nagging thought that there’s something missing here. Ten years ago, Greg Ciotti wrote that, “Great content is still the biggest hurdle for marketing teams” and I think that’s probably more true now than it was then. I’ve created a few hundred pieces of content just this summer, each checking every conceivable box that “good” content can check. On brand, on message, right length, format fits the channel, loaded with expertise, concise, etc, etc. My Claude Skills are really good (I know it because Claude told me so 😆). I can tell an LLM to make the content any way I want, but I still can’t force Claude to make sure it’s actually good. 

“Where focus goes, energy flows”

I was reviewing a batch of articles this week and reflecting on how my slick new system was producing work that looked the part but was lacking something that I can’t articulate well enough to bake into a Claude Skill. And perhaps, I realized, it is inarticulable. 

The above quote is attributed to Tony Robbins as best I can tell, but the idea is certainly not unique to him. It’s applicable here because the way we content marketers spend our time and energy is changing dramatically. In 2016, I could never have imagined that I’d be using AI to create content workflows and pushing changes to Github. (It’s wild to me how normal this has become in such a short amount of time.)

I realized that in my eagerness to use AI, I shifted my focus away from the one thing I am absolutely sure will make my work successful: making sure it’s actually good! It’s tedious and difficult and so obvious that you can miss it.

Let me try to be as specific as possible about this gap between content that checks boxes and content that is really good. I used to spend a lot of time making sure an idea was good before I wrote about it. To do this, I’d open up a new doc and brain dump. Then I’d do some research on the topic to see what else I could find. Had someone else already done it? Was there historical context that could inform my piece? Was there an existing framework I could borrow, or could I come up with a new one? Could I capture the idea in a single visual or chart? And on and on. Within a few hours, my doc would be a mess of links, snippets and sketches.

Eventually, I’d start some exploratory writing just to see if I could make the piece work. And if I couldn’t (something which happened all the time) I’d save the doc and try to use bits of it later. If it seemed viable, I’d begin that laborious process of turning that into something polished.

Writing great content is hard for (at least!) two reasons:

1. You have to vet your ideas with research, exploration, your knowledge of what readers’ want and your own taste. Author Malcom Gladwell captured this well on the Tim Ferris’ podcast, saying that, “To the extent that a writer deserves his or her paycheck, it comes down to how good are they at looking through a transcript and understanding what’s interesting. You have to sort of visualize what the story is going to become before it has become anything.”
2. You have to assemble the words in a way that’s easy and enjoyable to read, provokes thought and hopefully some discussion, and maybe even prompts the reader to take further action like joining a newsletter or sharing with a friend. 

The first step is understanding, the second step is knowledge synthesis. And in addition to being hard, each is time-consuming. I hear developers say something similar about writing code. It used to be so much harder and slower (not to mention expensive) that product teams were very careful about using the time. 

My content system abstracts the process but the whole is frustratingly less than the sum of the parts. It’s so fast that I often don’t bother tightly vetting an idea. Instead, I run it through my Claude Code workflow and see what comes out on the other end. I scan the draft and only decide then if I should start editing or adjust the input and try again. This feels wonderful at first because I’ve skipped the pain and suffering! But the more I do it, the more I see that skipping the friction results in a loss that can’t be fixed in a Claude Skill. The “Is this worth doing?” bottleneck collapsed and now we’re faced with a new problem.

After trying to avoid the hardest part of my job and only to find that the work is nowhere near as good, I would actually love to spend more time vetting ideas. But doing so actually feels counterintuitive these days. We are supposed to use AI for basically everything right, right? AI fluency is on performance reviews and taste isn’t. But I’m pushing back.

You are the driver, Claude is the engine

As I think about my work at Miro, I’m steering things back in the other direction. There are a few things I’m doing to ensure that the work we’re about to ship (more on this very cool thought leadership project coming soon) is good:

• Vetting ideas as if it’ll cost $10,000 to write the piece. It’s SO tempting to skip this but I’m certain my work will suffer if I don’t do this. Plus, writing based on a great idea is so much easier than editing a piece based on an idea that was only mediocre to begin with.
• Editing like it’s 2016. I am absolutely ruthless with editing content that has been written or even influenced by AI. I find this to also be difficult but satisfying work. I’ve got my copies of “On Writing Well” by William Zinsser and “100 Ways to Improve Your Writing” by Gary Provost sitting on my desk for the first time in a long time. (Also, for what it's worth, I have tried to bake some of these ideas into my Claude Skills with moderate success.)
• Collaborating! Sounds both obvious and lame, but I have thoughts on this. I think when the world embraced remote work, we also embraced asynchronous work. This was great because it killed a lot of pointless meetings, but I think we may have over-indexed on async. Claude is going to tell you every idea you have is amazing, but coworkers won’t. That friction leads to more creative thinking and better ideas. I am trying to spend more time outside the warm fuzzy confines of the “Claude Cocoon.”
• Leaning into the craft. I gravitated towards content working as a career because I enjoyed writing and was thrilled to find a way to earn a living. The craft of writing is still my favorite part about work and one that I'm leaning back into after spending a number of years focusing on other things. Even if much of the work is done by Claude, I believe that my own expertise in the craft of writing is vitally important. Many of us use AI heavily, but we cannot forget that we are in the driver’s seat.

Upon reflection, building Claude Code content systems (while a lot of fun) is partially an effort to avoid the pain and tedium of real editorial work. But I’ve come to the conclusion that it’s unavoidable and we should feel grateful for that. Writing is among the most rewarding tasks any of us spend our time on because it’s difficult. And besides, avoidance only accumulates pain.

I’m not abandoning Claude, not by any means. AI is part of our work now and I think we’re still in a phase where exploring how it can help us is worth doing. If nothing else, trying to explain to Claude what “good” means has forced me to articulate in a way I’ve never had to do before. What used to be reflexive is now well documented. 

Good content is hard because it has to be. Let’s embrace it. Writing suffers when humans don’t.