Hi everyone! My first post on dev.to. One of my many efforts to counter my timid nature.
That aside: 2 weeks ago, OpenAI quietly shipped Sites, a Codex plugin for ChatGPT Business and Enterprise. You can read the docs here: https://developers.openai.com/codex/sites. For those who don't want to read the long docs, here's the tldr: you describe a small tool in a prompt, and Codex builds it, saves a deployable version, and deploys it to a hosted URL. It can wire up a real database for saved records or object storage for file uploads if needed, but the apps themselves stay small and focused, closer to a single-purpose tool than a full product.
It's basically Claude's artifacts taken one step further: not a throwaway piece of JS/HTML living only in a chat thread, but a small, data-oriented app that can persist state and get a real URL.
The idea, i believe, is not to compete with app builders like Replit or Lovable, but to let teams turn a prompt directly into a small internal tool: a request tracker, an ops dashboard, a quick game with saved scores. Imagine needing a tool to triage requests from your team. Wouldn't it be nicer to just describe it and get a working tracker with saved state, instead of opening a no-code builder or briefing an engineer?
The idea is so cool. But then I realized I have something similar, and I can't wait to share it with you. Welcome to dim0.net, a canvas of things.
Dim0 takes a different angle: instead of a hosted deployment pipeline, the AI builds small interactive React apps, what we call mini-apps, directly as nodes on an infinite canvas. As you can see in the screenshot below, you can create rich notes, mindmaps, schemas, interactive charts, visual explainers, and these mini-apps side by side. You can ask the AI agent anything, and depending on the question, it will generate a mindmap, a schema, or a small interactive app right on the board, next to your notes.
It's a mix of Excalidraw and Notion. You can draw hand-drawn shapes and arrows, mix in rich-text notes (markdown underneath), and drop interactive elements wherever you need them, all wired to the context already on the board.
even better, dim0.net is fully open source (MIT license, repo here: github.com/vcmf/dim0), so feel free to check it out.
The engine behind the canvas is github.com/winlp4ever/canvas-harness, which comfortably supports thousands of nodes. It's a complex engine that deserves its own post.
That's it for now. I'd really appreciate any support here. Feel free to ask me anything, and thanks for letting me finally become part of this awesome community.
Top comments (0)