Stablemount is AI-first, and CMS second. The dashboard is primarily a browsing and prompting environment, allowing you to talk to AI directly inside the architecture, and audit/edit the results. The AI is built into the product — not an external coding tool you connect. A marketer uses it the same way a developer does.

This is not vibe coding. Tools like Cursor and Copilot are powerful, but they're AI baked into the IDE — they help developers write code faster. Stablemount is AI baked into the product architecture. The person building your about page doesn't need to know what HTML is. They describe what they want inside the dashboard, the system builds it within the site's existing design system, and they edit their own words on the rendered page. The constraints — design tokens, instruction rules, reusable components — live in the architecture itself and get denser with every page built, not in a developer's repo context or cursor rules file.

There is no separation between content and editing. No TinyMCE, no Portable Text. Instead, Stablemount converts live HTML components to contenteditable, allowing you to edit, update, and alter site content directly on your front-end. There is no traditional visual builder system — there is no need for one. The AI is the visual builder.

There are no dependencies aside from the Anthropic SDK. The architecture allows the AI to write flat HTML/CSS/JS files. Individual partials in your components library are all single HTML files. Want to copy a component between projects? Copy and paste the single file. Everything is git-diffable, portable flat files on disk. MIT licensed.

I think people are still viewing AI as something that "will integrate into their current workflow." In my opinion, the truth is that AI will redefine an AI-first workflow from step 0 to release. Every CMS is going to be AI-first within 18 months. The question is whether they rebuild the architecture or bolt AI onto what exists.

Comments URL: https://news.ycombinator.com/item?id=47650854

Points: 3

# Comments: 1

That may be a bit of an exaggerated turn of phrase, but I believe its actually fundamentally true. Language has always been a human tool. It has always been a tool for human-to-human communication. The only things in the universe capable of properly, genuinely using human language were humans.

That is no longer true, and because the mutation started as a silly little chatbot, and because the mutation has been cloaked within the promise of "increased productivity" and other various obfuscated terminology, this shift in the functionality of language has gone largely unnoticed.

I would argue that even without the enormous future possibilities of AI coming to bear, the technology has nevertheless already drastically altered a key facet of human existence. Instead of human language, we now have "cognitive language," a necessarily broader term to account for the fact that its not just humans using it anymore.

I think this is a more impactful change in society, irrespective of other changes, than people are giving it credit for. Honestly, I don't think most people have really internalized the implications (including myself), and I don't see anyone discussing this change in the broader conversation around AI.

Comments URL: https://news.ycombinator.com/item?id=47642468

Points: 4

# Comments: 1

I've built Wordpress sites for 12 years. Very few Wordpress developers are trying to swap to a slightly upgraded version of the same thing with no ecosystem and much of the same solutions. This will see some adoption, no doubt, but not a serious dent.

The main reason for that: in 12 months, 24 months, 36 months, this solution will be outdated and unimportant, same as Wordpress. Wordpress will still be kicking because it already has a 40% market share on the entire internet. This, however, will not be.

The CMS is dead tech in six to twelve months. I might have a million people who will disagree with me (and yes, people will still use CMS's after twelve months), but people actually moving into the future will have dropped CMS's for architectures that are AI first with strong, intuitive, easy-to-leverage guardrails.

In my opinion, the vast majority of people are still looking at AI through the lens of "how does it alter my current work/tech stack/strategy" and failing to ask the proper question: "what the hell is even important in a world where AI is as competent as 90% of humans and 100x faster?"

What do you need a CMS for? You think you'll be managing the content? Why? Why do I want a human managing content when the AI does it 100x as fast? Why do I want Astro? It compiles down nicely? Okay, maybe its a god-tier solution, but more likely... AI can just code extremely fast vanilla html/css/js. Why do I need a component library when AI can steal all the best components from all the best libraries? Why do I need "Portable Text"?

This is still not big picture enough. Think further out than 12 months.

Writermark collects telemetric (and anonymous) data about keypresses, records your total time spent on a document, validates that the text has not been altered inorganically or pasted into the editor, and scores the final result. If the score is high enough, a certificate is issued that can be verified at https://writermark.org/verify at any later time. The server is stateless and stores zero information -- all your data lives on your local machine in a server-signed JWT.

Try it live: https://writermark.org/#try or integrate it into your own app: https://writermark.org/developers (npm install @writermark/sdk)

Or download Wintertext (https://wintertext.com), a free desktop writing app with Writermark built in. Wintertext takes this ecosystem a step further, packing the certificate and the text together into a .wtxt proprietary file type, while also allowing export and verification for .docx, .md, and .rtf files. Any document created with Wintertext is automatically certified, lives on your local machine, and the app will remain free forever.

I built Writermark because I think we're losing the ability to tell whether a human wrote something (which scares me deeply, especially in the world of fiction). If you are also concerned about the progress of AI infringement into human writing, I'd love to hear from you, hear your feedback, respond to any questions, and start a larger conversation.

If you're curious to know more about the general philosophy, feel free to read this longer overview on the tech: https://www.writermark.org/blog/working-to-verify-and-protec...

I am currently in the process of establishing nonprofit status for this project. I think that human authorship verification should be free. Whether its my solution or someone else's, I think telemetric verification of the writing process is going to become one of multiple important ways that we prove that something is human authored.

Thank you very much for your time.

Yolm.

Comments URL: https://news.ycombinator.com/item?id=47478224

Points: 2

# Comments: 0

No information about what threshold I need to cross, what the requirements are, what I need to do to post my project.

Very cool.