Honest question. Do people read these massive vibe-coded README.md files, or are they just artifacts?
The minute I see them, I tune out and move on. These types of docs come across to me as so over-specified and memorialized that I just find it not worth the effort to read. Of course for legacy projects this type of documentation would have been invaluable. What's changed, I wonder?
If vibey readmes are for LLMs like the other comments suggest, they should be banished to a folder where it's clear they're just for a robot so I don't have to look at them.
The README.md can actually be a reserved space for human beings to read with whatever the author thinks is most important to communicate to the human operator.
Claude.md serves that purpose by standard and does not edit readme.md unless prompted so. I would keep it like that (or use another name for different agents).
My eyes just glaze over when I see every file being listed and its purpose. Just because you CAN do that with LLMs doesn’t mean you should. What are you really trying to communicate?
I was under the impression the purpose of that is to give an LLM a cheap (in terms of context window) understanding of a repo, and not at all for the benefit of humans?
Whether that's effective is another matter, even if the LLM generating the list does so correctly and updates the list consistently
If we're fair, this is one of the better examples. It looked reasonably useful until it got to the 'CI / GitHub Actions' section. Many readmes take a very long time to describe anything useful, like how to get started or even what exactly the software is.
README should be written for humans. Agents don't need fancy rendering in the repository page, we do. Agents already have their own dedicated context file.
Yes, they are typically pretty useful for reproducing the setup. I argue also they are superior to most human made README.md because they explain what the project is at an abstract level, use more features of GitHub Markdown to communicate, and also generate diagrams which are helpful to visualize concepts.
You have. The words in the file still convey useful information, but because theres a particular writing style, you've deemed the words unimportant to read. If you went to a lecture about some topic you were interested in, but as soon as the speaker started talking, they sounded gay, and then you got up to leave because you just can't stand that, is that on you or them? Because the AI culture war of it's coming to take your job, so you'd better detect when something is made by AI, otherwise you can't enjoy it. Calling out every instance where AI is slightly off is a good thing to do. Because it'll be used to make the machines better at not doing that.
It's not only slop, it's low-effort slop. How does you post a link including "in the Browser", but actually running it requires some vague pointers about spinning up docker on a remote x86 linux machine?
But... why.... from all possible protocols and options out there for secure, and less complex channels out there... they chose the so featureful and pure overhead as SSH...
For low-latency VNC+Audio, why not WireGuard, WebRTC, QUIC etc.?
Yes, Dungeon Crawl Stone Soup approached it relatively similarly for WebTiles, yet... it's infinitely different magnificent titles...
Oh... it's not them... it's Claude they used... or? Who knows, right? No one, including the priceless invaluable ideas of actual developers meat-ground into the Claude datasets now being sold for tokens...
// https://i.imgur.com/FM4aTMZ.png - SSH
// https://i.imgur.com/9lZniqE.png - Claude
Sooo it’s just streaming a game to your browser. What does this have to do with Dwarf Fortress? There’s tons of generic solutions like this that would work the same way for DF as well. It’s giving AI slop.
This is the sort of trademark violation where I can get behind proper enforcement. It's using the name of an unrelated project for advertisement (attention).
Play the classic (ASCII/2D) Linux build of Dwarf Fortress in a web browser. DF runs as a Docker container on a remote x86-64 Linux host at full native speed and is streamed to your browser over noVNC.
The minute I see them, I tune out and move on. These types of docs come across to me as so over-specified and memorialized that I just find it not worth the effort to read. Of course for legacy projects this type of documentation would have been invaluable. What's changed, I wonder?
The README.md can actually be a reserved space for human beings to read with whatever the author thinks is most important to communicate to the human operator.
I've seen legacy applications with a case statement that spanned 30,000 lines. Apparently human coding at it's finest.
Whether that's effective is another matter, even if the LLM generating the list does so correctly and updates the list consistently
This is just yet another low-effort slop specimen.
Coke Zero is good these days.
You keep posting this. What's it for? Who wants this? What does it accomplish?
It's a vibe-coded remote console renderer. I don't get why you're posting it.