The human-first CMS

This is the interface of a data entry clerk. Headless architecture, visual editors, structured content models: real progress, but progress within the same paradigm. The editor fills in fields. The system manages what gets filled in.

We've published three posts this month, each starting from a different question. Reading them back, I realize they were all circling the same thing.

Three angles, one observation

The first was about the blank page. Marketing teams open a polished editor and freeze, because it gives them a perfect place to write and zero help knowing what to write.

The second was about the editor as the last mile. The "agentic CMS" conversation is stuck on infrastructure, but the part that determines whether agents produce anything worth publishing is the editorial layer.

The third was Anders writing about ARC-AGI-3. Machines still can't do what humans do on first contact: walk into something unfamiliar and make sense of it. The same kind of judgment an editor uses when deciding what's worth saying and whether a draft has found its voice.

Three angles. Same observation: the part of the work that matters most is human judgment. And the CMS isn't designed for it.

What they were circling

AI accelerates the paradigm without questioning it. Agents generate text for the body field. They suggest tags, optimize metadata, translate. The pitch is "agentic CMS": do more with less. Faster data entry is still data entry.

The blank page post was asking: what if the CMS helped you figure out what's worth saying? The last mile post was asking: what if the editor's job was directing, not producing? Anders was asking: what if the capabilities that matter most are the ones machines can't replicate?

They were all asking the same question. What if the CMS was designed for taste?

What we're building

My own workflow changed this year. I used to open an editor and start typing. Now I describe the angle I want to explore and the audience I'm writing for. Agents research the landscape, structure the outline, draft against a voice guide, run a review, and prepare social posts.

This morning I used that workflow to write a post about the CMS and editorial judgment. The agents did their job. The draft was clean, well-structured, voice-compliant. I read it back and killed it, because it said the same thing I'd already said in the last mile post. The angle was right. The argument was redundant. No review agent flagged that. Only taste did.

I write less than I used to. I think harder about what I want to say and why it matters. My job is taste.

That shift is what we're building into Strife. Most CMSs treat the editor as a data entry clerk. We're designing for the role that actually matters: the creative director. The person who sets editorial direction, provides the brand's why, and makes the judgment calls that only a human can make. We've started calling it a human-first CMS, because that's the design principle: every interface decision starts with the question of where human judgment adds the most value.

The industry is adding AI to fill in fields faster. I think the more interesting question is what a CMS should ask you to do instead.