A Well Thought Out Harness Can Substitute Anything
It doesn't need a new interface built around it. It just needs a good enough one already doing the thinking.
Most interfaces exist for one reason: something needs to get from an idea to a system, and a human has to be the one carrying it across. A form, a panel, a menu of buttons, those are all just different shapes for the same job. The click isn't the point. The click is what stands in for a decision that already got made somewhere else.
Once a harness is capable enough, and trusted enough, that carrying job stops needing a human, and this isn't a special claim about any one tool. It's a general one. A well thought out harness can substitute for almost anything standing between an idea and its execution: the interface that ships a build, moves a file, sends a message, or turns a finished draft into something live. The task changes. The pattern doesn't.
I've started testing that claim against my own setup, one interface at a time, and the one it's up against right now is Payload, the CMS that runs the backend of this site. It's where every post, image, and field on tumaini.dk actually lives, and until now, the admin panel was the only door into it.
Every time I've had a finished draft, I opened that admin panel, found the right collection, pasted the text in, set the fields, and clicked publish. None of that is hard. It's a series of clicks standing between "this is done" and "this is live," and clicks like that are exactly the kind of thing a harness is good at absorbing. So the harness gets direct publish access. I hand it a finished draft the same way I'd hand it anything else. It writes to the same posts collection through the same API a manual entry would use, sets publishedAt, and the post goes live. Nothing about the site changes. The only thing that moves is which interface says "ship it."
Insight: A well thought out harness can substitute for almost anything standing between an idea and its execution. It doesn't need a new interface built around it. It just needs a creative thought to happen, and it takes the rest from there.
Payload isn't going anywhere, and that's worth saying plainly, because the point here was never to get rid of it. It's still where I go if I want to hand-edit something, preview a post before it ships, or fix something from my phone. It's a good tool, and a good admin panel is worth keeping around. It just doesn't need to be the only door in, and once you notice that about one tool, you start noticing it about the rest of them too. The form was never doing the thinking. It was just where the thinking got typed in.
That's the actual shift, and Payload only happens to be the tool I'm testing it against first. The next interface that gets absorbed probably isn't a CMS. It's whatever else in my day still assumes a human has to be the one pushing the button, once the harness already knows what the button is for.
I keep finding places where a well thought out harness quietly replaces something I assumed needed a human. The skills and workflows I build around that idea live at github.com/Kimotep/skills.