A short defense of writing where you work.
On chatbots as composition surfaces, the death of the CMS, and why the most honest writing happens before you’ve found the perfect adjective.
For twenty years, “publishing” has meant opening a different application. You finish your sprint, you close your editor, you find the CMS tab, you fight the rich-text editor, you re-find the link, you choose the slug, you upload the photo, you remember the photo’s alt text, you preview it, it looks wrong, you switch back to figure out what changed, you’ve now lost the thread of the post itself. By the time it ships, the moment has cooled.
This is not a problem with willpower. It is a problem with surface area. Every CMS is, in the end, a context switch — a deliberate exit from the place where the work happened, into a place built for managing it.
“Most people don’t fail to publish because they have nothing to say. They fail because saying it requires them to go somewhere they don’t want to be.”
The chatbot is the composition surface
Look at where serious thinking happens in 2026. Not in Notion. Not in Word. Not in WordPress. It happens in a chat with Claude, or ChatGPT, or whatever your team has standardized on. That is where the loose drafts live, the half-formed arguments, the explanations of what you just did. The chatbot is already a writing tool. It is, secretly, the best one we have ever had.
scsiwyg starts from that observation and asks: what if your chat could publish? Not by exporting Markdown. Not by copying out a result. By calling a tool — speaking the verb publish the way you would say git commit — and having something appear, on the open web, with a URL, in your voice, under your own domain.
The editor is dead. Long live the editor.
scsiwyg has no CMS. There is no rich-text box, no media library, no draft list, no formatting toolbar with the bullet button next to the numbered-list button. There is an API. Talking to that API is the entire interface. The agent is the editor; you are the editor’s editor.
This sounds austere. It is, in practice, the opposite. You don’t have to learn a new layout. You don’t have to remember which dropdown contains “Schedule.” You don’t have to fight a WYSIWYG that thinks it knows better than you do. You write in the place you already write — your editor, your terminal, your chat — and a small piece of plumbing turns words into pages.
Who this is for
scsiwyg started in the developer ecosystem. The first users were engineers using Claude Code or Cursor who wanted to keep a working journal of what they shipped. Those users are still here, and they are still the spine.
But over the past year we have watched a different population arrive. Designers logging the rationale behind a decision before they forget it. Researchers maintaining a living wiki of papers they’ve read. Founders publishing a build log to twelve people who care. Anyone who already lives inside a chatbot — and increasingly, that is everyone — has a publishing surface, now.
“If you can ask Claude a question, you can run a magazine.”
What stays human
The agent does not publish anything you have not approved. There is a hard step between draft and live; it is, deliberately, the only friction we kept. Everything else gives way: the slug-naming, the tag-picking, the markdown-fighting, the scheduling. What you do is decide whether the thing reads like you. Then you say publish, and it does.
This is the small wager scsiwyg is built around: that the only step worth defending in publishing software is the part where a human reads what’s about to go out and decides, yes, that’s the version I’d defend in public. Everything before that is plumbing. We’ve taken the plumbing as our problem, so the writing can be yours.