Pre LLM nodejs was convenient option to deploy to all platforms at once.
Post LLM that's no longer a limitation! Just prompt more to create apps in native language! Anyways RAM is getting costlier by the day due to LLMs themselves
That is unless devs don't use LLMs at all. Then they should simply state in the GitHub repo that we don't use AI!
I wouldn't use AI for hobby projects tbh. It's my personal preference
If you have ever done that you know that that's as true as "write once, run everywhere". There will always be tweaks and iterations, even if it saves you a lot of time.
Since this project uses such good agentic engineering practices, I was able to use Claude Code to add support for Mermaid diagram rendering and produce my very own local build on Mac, all in about 30 minutes.
Thank you! Hubble+cmux gets me close to the agentic IDE I’m working towards.
Can you easily share lists or pages or is it all local? (Is there a cloud element to it)
What are the most useful examples you’ve personally found when it comes to using the HYML component?
I built https://smalldocs.org for my agent’s Markdown needs. I extended it so it can render complex formats: Mermaid diagrams, charts, in-document slides and even in-document spreadsheets. It gives your agent a good selection of formats to express itself as it helps you understand whatever topic you’re looking into.
Here is a sample document: https://smalldocs.org/blogs/what-is-a-smalldoc
I publish anonymous analytics here: https://smalldocs.org/analytics
I can let my agent edit those files all the same + there's better integrations and knowledge readily available on these.
I can tell my agent to edit the Obsidian Markdown just like that, but I can also tell it to create a full-on notebook with backlinks and DataView queries. I can tell my agent to create logically complex .org notes with code blocks and references.
To me, Hubble just seems like a Markdown editor, based on the premise that agents are able to edit local files.
Now, I'm having fun with tldraw.[1] With their new tldraw offline, I can continue to work in Claude Code (and others) and I can either output or interact/input from tldraw. This is more of a free-form, infinite canvas. I treat it like Excalidraw + Draw.io + Freeform Notetaker.
For text, most agents/LLM/assistants are pretty happy and smart using the same Markdown files to anywhere I have allowed them to interact.
It has a rest api and a cli tool, and also the possibility to make headless deployments, and it can sync clips without a sync server. It can serve folders/tags and optionally expose a read or read/write api for JSON files on the same level. It supports lua plugins, and I have a fal.ai plugin to generate/edit images already in the repo (some other examples there).
It has a lot of features so maybe it's better if I just link to the docs: https://egeozcan.github.io/mahpastes/
I feel like tables are generally underrated while the AI ecosystem is focused on .md files but I guess markdown has tables so they’re included to some extent