Most content advice assumes a team. An editor, a scheduler, a designer, someone who handles distribution while you focus on writing. The advice sounds reasonable until you’re the only person in the room.
I publish four articles a week on Freymwork. I write Substack Notes on top of that. I have client work running in parallel. I’m not telling you this to impress you – I’m telling you because the system I use is shaped entirely by that constraint. One person, limited hours, no room for fragile workflows.
Here’s what actually runs it.
It starts with a theme, not a title
Before I write anything, I plan. Not individual articles – cycles. A group of articles built around a central theme.
The reason is simple: if I write about topic A this week and topic B next week, I’m always starting from scratch. New angle, new context, new research. That’s fine when you have the bandwidth. I don’t. More importantly, neither do my readers. A cluster of articles on the same theme gives them something to go deeper into, not just a series of isolated pieces.
Planning happens in two phases. I use Claude to brainstorm – what belongs in this cycle, what angles are worth exploring, what I might be missing. I run it as a conversation. I push back. I redirect. The output is rough material, not a final plan.
Then I take it into Craft and do the real work by hand. I edit the structure, cut what doesn’t hold up, add what the AI didn’t see. The plan document becomes the foundation for the entire cycle – the reference point I come back to when I’m mid-article and need to remember what the whole thing is actually for.
What makes the drafts usable
This took longer to figure out than I’d like to admit.
I spent a significant stretch of time trying to make Claude generate drafts I could use without rewriting them from scratch. It didn’t work because the context wasn’t there. Generic prompts produce generic output. I was getting text that was technically fine and completely not mine, and worst of all they felt AI-written. AI-slop is not what I want to publish, nor something I want to be known for.
The fix was building a dedicated Claude project for Freymwork. It contains background on what the project is and why it exists, context about my experience and perspective, and a content system that governs how articles get structured. When I prompt for a draft now, the model already knows the voice, the format, the principles, and the audience.
The drafts still need editing. They’re not finished. But they’re a real starting point – close enough that my editing time is spent refining, not rewriting.
The actual workflow, step by step
Once the plan is in place, execution follows a consistent sequence.
I write in two batches per week – two or three articles per session. Each session, I prompt Claude for a draft using a structured brief that’s part of the content system. That brief also includes a request for what I call “publishing support” – suggested tags and a description I use as instructions for image generation. Image is generated separately in Gemini via a custom setup. Everything comes out of one prompt, which matters more than it sounds. Fewer separate decisions, fewer context switches.
The draft goes into Craft for editing. That’s where I read it properly, slow it down, and make it sound like me. Craft is where the plan lives too, so I can cross-reference without jumping between apps.
After editing, it moves to Substack. Substack Notes are scheduled through a custom WordPress plugin I built before Substack had native scheduling – and kept using because it’s faster and fits better into how I already work.
The whole process – four articles plus Notes – takes around four to five hours a week.
The constraint that shaped everything
I don’t want to context-switch.
Because every time you shift context, you pay a tax. You lose the thread of what you were doing. You have to rebuild it. In a team operation, you can distribute that cost across people. Alone, it compounds.
The theme-based planning reduces it at the idea level. The Claude project reduces it at the writing level. Craft as the central hub reduces it at the editing and storage level. The plugin reduces it at the distribution level.
Every tool in the stack exists because it removes a decision or eliminates a switch, not because it’s the most sophisticated option available.
What I still do slowly on purpose
I could publish faster. I choose not to.
Editing is where quality actually happens, and editing takes time. Not because the drafts are bad – they’re genuinely useful now – but because I want the finished piece to be sound. True to what I actually think. Useful in the way I mean it to be useful, not just technically correct.
Four articles a week at this quality level is the ceiling I’ve set, not the floor I’m trying to escape. A system that produces more at lower quality isn’t an upgrade.
The point of the system isn’t speed. It’s consistency – the kind that holds up on a Thursday when energy is low and the last thing you want to do is figure out where to start.
