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.
Yeah I was expecting some giant text dump based on the complaints in the comments here, yet all I see is a pretty normal README? It's not even that large.
If we are splitting hairs some sections could have been left out or broken up into a dedicated doc, but this is still vastly better than a README that explains too little. At minimum I expect a README to give me guidance on what this is and how to get started with it and having stuff like a file layout helps with exactly that.
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.
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
Yes, that's for the computers to read. Maybe it's documentation that might be useful for a developer new to the project?
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.
Yeah I was expecting some giant text dump based on the complaints in the comments here, yet all I see is a pretty normal README? It's not even that large.
If we are splitting hairs some sections could have been left out or broken up into a dedicated doc, but this is still vastly better than a README that explains too little. At minimum I expect a README to give me guidance on what this is and how to get started with it and having stuff like a file layout helps with exactly that.
We're in an era where the documentation is written to be read by LLM's, and I fully expect the README files to follow suit.
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.
Isn't this the whole point of an AGENTS.md file? So that you can leave the readme for humans and put LLM-friendly crap in a separate file?
If I see an excessively long README which is competing to being as long as War and Peace, I am immediately not going to read it.
This is just yet another low-effort slop specimen.
Are we just saying this now for everything? The readme seems fine and is not long. The entire file is scannable in 1 minute.
It's 170 lines. that's not "massive".
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.
Coke Zero is good these days.
its less the tone and more the oppressive volume of text
Why?
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.
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?
Isn't it kind of just as easy to stream it from your desktop into your browser? Something like RustDesk doing remote desktop?
Geforce Now has had Dwarf Fortress for quite some time also...
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).
Could this work with the actual game files (e.g. purchased from Itch.io instead of Steam) ?
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.