| Think of it as a cross between Google Docs, Gist and LLM chat (with tools). What it does today:
Editor for GitHub-flavoured Markdown + Mermaid + KaTeX in one place
Live preview as you type
LLM Chat enables:
Drafting an architecture doc from a short description
Generate or update Mermaid diagrams from text
Refining a document Change Tracking (visual or diff)
Public share links Integrations (needs tidy up/fixing) ImageGen - hooked up to flux.1, gen image store in s3 GithubMCP - needs fixing (too many tools/ no private repo support) Web Search/Fetch How I got here (14M later):
It started out as a development platform with vscode plugin and support for a frontend for docs/design with realtime comms between the two. Decided this was a losing battle as a side gig against cline / cursor etc . This still exists, just isn't exposed and is what I've used to develop AutEng - it's very specialised to my workflow (documents for arch -> phased plan -> progress log).
I switched to Goal Driven Agent, where the LLM generates a json-schema digraph to meet the goal. This was difficult (it works, just not very well), but I still have a workflow automation I can use down the line (e.g. auto create docs from github).
Third and hopefully final pivot to Docs, I was struggling sharing DS conversations and ML architecture proposals I'd had with various LLM's and realised AutEng was capable of this. What’s next Kind of depends how things go. Directions I'm thinking: Proper Github integration / autogen doc from repo Shared realtime editing (Google docs style) Support for more niche rendering e.g. chemistry, biology Check out where I am (LLM Chat needs signup, but still free for a try):
https://auteng.ai (loads straight into the editor, no login).
Example architecture doc: https://auteng.ai/docs/software/architecture/architecture-do...
Example math/physics doc: https://auteng.ai/docs/physics/general/physics-equations-dia... Feedback would be very much appreciated, particularly: What would this need to be useful to you?
For math/physics folks: is this usable for real derivations/problem sets, or what’s missing?
For the autogen docs and keep them up-to-date, how would you want that to work (e.g. PR to github/gitlab repo) Happy to answer anything about the journey, product, roadmap etc. |