Yeah, Reading it clearly in the editor - Probably MD to HTML happens in the editor anyway. So what we are dealing with here are a battle between ease of writing, readability and various cost dimensions.
If cost of showing HTML and MD are same in the CPU level for UI, but generation of them costs uneven at AI level (as it should be mathematically), shouldn't the equation be what is the lowest cost to represent something considering the token usage
A) Enterprises for one, e.g Rendering of an HTML page vs MD in an AI IDE or even tokens (compounding costs for a communication doc may be billions in a year)
B) True, its a subset, nobody is arguing against it. HTML has overheads in terms of stylings. MD in essence is just a straight arrow.
If there is a communication between two points with ultimate low latency way with readability and threshold just enough to have the idea, MD passes that mark. As tokens needs to conserved
Thats a complete reductive idea of MDs. MD was long spawn before LLMs as a need for light weight formatting and easy conversion to different formats. No one used at that scale doesnt mean the absence of the utility for it, as the primary adopters were tech community and that was a layer below the consumer market. Utility obviously stands, the question is, can HTML supersede on that utility side of it in all developer work components.
With AI IDEs, Personally I had to generate a tons of md files for planning a task, analysis on the code for something or other, a task doc for a feature - a summary to paste into the clickup ... and I saw many devs keeps on generating them, and all tucked away into folders. Good, okay.
For the company I'm currently working I had made a VSCode extension where I can sync the task doc with clickup via frontmatter.
I decided to take it to next level as a side project. I built a CI integrated, git-native, agent template transformable syncing pipeline with git MD files to any project management tools. That means, either you can save your md files vanilla in your wiki (thus using the clickup AI search to dig up later, get insights etc) or you can use a AI agent template transformer to turn it into a task template (Background, acceptance criteria, functional requirements etc.) and update or create a task on a board.
I've been working on it now. I don't know how it will fare, but I feel like product is coming up nice.
There was a plug of smalldocs intro in that post, regarding that,
Dont we need the md files on the wikis and documentations in the central project management tools, instead of client side anything?
Yes, In my experience with it, AI agentic syntaxes were few notches above what human would do, where in humans quality can fluctuate, and as creatures of habit we all tend to develop some patterns that can be read by our reviewers intuitively when our name is on it. AI can consistently get the right patterns if we can keep a tight rule/skill/checks on them, but thats mean they would conform to one pattern.
imo, unless you can quickly read off the approach the AI has taken, If you haven't got the experience for it, it can be a minefield. Because by agentic coding, we have moved beyond syntaxes into the ideas and approaches. Agent gives an approach, we should know if its the efficient one. But if you got that reading hook, its really fast and I dont see why it should not be production code.
true, yeah thats right, But my larger point is we are operating many levels above than we can with JS. Like C, C++/Rust is the closest as LLMs would be trained on that.
I just threw the extreme point on the curve, to see where we can land on.
Yes, front end is indeed in the automation territory. Artisanal works aside, common patterns usually serves peoples needs, so vibe coded front end will in fact work.