rns.recipes

Due to the influx of LLM assisted shovelware, these are the rules that are now required if you use LLMs either in part or full for development to promote your project here.

• You must demonstrate that you have a clear knowledge of your project's codebase.

This one is very simple. The most immediate sign that a project is LLM shovelware is if you can't even bother to write your own README. If I see — a slew of — em dashes everywhere, "It's not ABC — it's X, Y, and Z with DEF", 🚀 emoji and table vomit that LLMs love, and the usual slop, I can more or less gleam in 30 seconds your project is shovelware.

This is embarrassing. There are even ways to deslop LLM output if you can be assed to put in the bare minimum effort. https://github.com/realrossmanngroup/no_ai_slop_writing_rules

Another one is clear functional design patterns.

A common design pattern used by LLMs that makes things borderline malicious is excessive announces. The number of projects spewed out by LLMs that love to try announcing every 10 seconds and for some bizzare reason shoving everything in the app_data of an announce I can't even count.

If your project is a LLM assisted port of the Reticulum Python implementation, it must be compatible with the Python reference implementation. If it can't even work with one of the scripts from the Examples folder then go wrangle your LLM until it does. There are like 6 LLM tooled Rust "ports" sitting on Github that can barely handle a link, despite reporting "feature complete".

The next one:

• Your project must have 6 months of commit history

Most LLM projects I see particularly in relation to Reticulum are abandoned after the first few months with little support afterwards. 6 months of commit history is more than enough to demonstrate that you are committed to maintaining your project, even if that is through the use of LLMs exclusively.

And lastly:

• Your project or post should have an LLM disclaimer usage. This is not a requirement but a recommendation.

That's it.

As the author ("editor"? "curator"?) of an LLM-assisted Reticulum port, I find these rules fair.

Of course a bummer that this means reticulum-js won't be listed here in next five months, and instead the broken-and-unmaintained rns.js is.

But at the same time, the five months will give me time to get much further with the port. Now we're pretty good with running a "leaf node", but I do want to get transport node capabilities in. And that also means looking at all those new routing improvements that landed in Python rns 1.4. And in that time I should have a few more reticulum-js powered projects up and running on top of the existing Signal K and Yjs integrations.

In that timescale we will also be sheltering from the southern Pacific cyclone season, which means I will have at least a few weeks of dedicated time to work on this instead of sailing around.

bergie wrote:

As the author ("editor"? "curator"?) of an LLM-assisted Reticulum port, I find these rules fair.

Of course a bummer that this means reticulum-js won't be listed here in next five months, and instead the broken-and-unmaintained rns.js is.

But at the same time, the five months will give me time to get much further with the port. Now we're pretty good with running a "leaf node", but I do want to get transport node capabilities in. And that also means looking at all those new routing improvements that landed in Python rns 1.4. And in that time I should have a few more reticulum-js powered projects up and running on top of the existing Signal K and Yjs integrations.

In that timescale we will also be sheltering from the southern Pacific cyclone season, which means I will have at least a few weeks of dedicated time to work on this instead of sailing around.

Yes its fine you can post updates about reticulum-js no worries

This is why global ret blackhole management is a must against the ever increasing shovelware across transport nodes. Hyperscaling agents do not respect the Zen manifesto (as bandwidth is finite).

A little while ago, I posted something saying "this is why we need Reticulum" and referred to censorship in Russia. More than one person gave me their "how dare you unilaterally declare the purpose of Reticulum"

Ironic, since I could quote the "Zen of Reticulum" first page of text in support of the comment. Also, I wasn't excluding anyone else's use of Reticulum by making my statement

Here in this post is someone unilaterally deciding what other people are or aren't allowed to do. Along with weird pet peeves that don't actually say anything about quality.

For example, supposing that an ESL person decided to use AI to write his README for better grammar and clarity. Each paragraph, he writes his version of the description, and AI translates it into better English. Is he going to then edit out the em dashes? No, because they're irrelevant to the meaning of the paragraph.

By all means, provide your support or opposition. Better yet, justify it concretely. But if you don't think that a human can spam announces or have half written poorly maintained and buggy code, you are sorely mistaken.

Reticulum assumes that every peer on the network is potentially hostile, and every link is potentially compromised. You're not going to be able to shame people into not being compromised. The one thing that nobody can beat, no matter how much AI is used, is physics. You're not doing anyone any favors by trying to pretend that we're not operating at that level

jrl290 wrote:

A little while ago, I posted something saying "this is why we need Reticulum" and referred to censorship in Russia. More than one person gave me their "how dare you unilaterally declare the purpose of Reticulum"

Ironic, since I could quote the "Zen of Reticulum" first page of text in support of the comment. Also, I wasn't excluding anyone else's use of Reticulum by making my statement

Here in this post is someone unilaterally deciding what other people are or aren't allowed to do. Along with weird pet peeves that don't actually say anything about quality.

For example, supposing that an ESL person decided to use AI to write his README for better grammar and clarity. Each paragraph, he writes his version of the description, and AI translates it into better English. Is he going to then edit out the em dashes? No, because they're irrelevant to the meaning of the paragraph.

By all means, provide your support or opposition. Better yet, justify it concretely. But if you don't think that a human can spam announces or have half written poorly maintained and buggy code, you are sorely mistaken.

Reticulum assumes that every peer on the network is potentially hostile, and every link is potentially compromised. You're not going to be able to shame people into not being compromised. The one thing that nobody can beat, no matter how much AI is used, is physics. You're not doing anyone any favors by trying to pretend that we're not operating at that level

I never once, in any of my posts condemned the use of LLMs entirely.

I am not going to answer every gotcha or trick question like I am in some court of law. If you want to make software it has to work. Not pass some random shell script invention "Harness" that your "agent" shat out. It has to WORK in the objective meaning of the word.

Thanks for reminding me to delete your shit too, btw. Your "RFed" sloppulus and "mobile app" DON'T FUCKING WORK. A nice usable MTU of 70 bytes to maybe fit a handful of words in and a jank broken IOS app that does about 1/2 of what is even needed for minimum viable functionality. All shat out by an LLM of course. Because you have literally no understanding of anything and rely on the black box to do it for you.

If LLMs could produce WORKING, FUNCTIONAL code, I wouldn't give a shit. I literally would not care. I'm not some luddite against LLMs., I am against people shitting up this forum with slop. And if you have to ask "is this slop" the answer is probably yes.

I never once, in any of my posts condemned the use of LLMs entirely.

Listen to yourself. You're not even responding to something I said. Just look how infuriated you get by the notion that em dashes aren't a good indicator of code quality

Your "RFed" sloppulus and "mobile app" DON'T FUCKING WORK

They work for many things. And for other things they are still being developed. That's how development works. Things don't work. And then they do a little bit. And then something else goes wrong. And so you develop some more. And you get some people to help test for you. Sometimes that even happens within 6 months of publishing on a repository...

The sad thing is, the people you're railing against, that you hate so much, they aren't in these forums. Nobody here is DoSing those backbone nodes, on purpose or otherwise. You're so angry. You have to shit on the people looking to help build something ("luddite" is written all over that comment btw)

Anyway, I'm sure that these social skills have worked awesome for you in life. I'll keep on writing code that nobody else is. And I'll happily spend my time, effort, skill, and money to do so

jrl290 wrote:

I never once, in any of my posts condemned the use of LLMs entirely.

Listen to yourself. You're not even responding to something I said. Just look how infuriated you get by the notion that em dashes aren't a good indicator of code quality

Your "RFed" sloppulus and "mobile app" DON'T FUCKING WORK

They work for many things. And for other things they are still being developed. That's how development works. Things don't work. And then they do a little bit. And then something else goes wrong. And so you develop some more. And you get some people to help test for you. Sometimes that even happens within 6 months of publishing on a repository...

The sad thing is, the people you're railing against, that you hate so much, they aren't in these forums. Nobody here is DoSing those backbone nodes, on purpose or otherwise. You're so angry. You have to shit on the people looking to help build something ("luddite" is written all over that comment btw)

Anyway, I'm sure that these social skills have worked awesome for you in life. I'll keep on writing code that nobody else is. And I'll happily spend my time, effort, skill, and money to do so

Blow it out your ass.

Btw, anyone else reading this who thinks this is harsh or mean to not just relentlessly shit up this forum with stuff that doesn't work cranked out in an hour with Claude Code can fuck off too. Maybe Reddit is more your cup of tea. Thanks!

Oh yeah, "writing"? Yeah, lot's of writing there bro.

Oh yeah, "writing"? Yeah, lot's of writing there bro.
[image]
[image]

That image shows a 5 month old repo with Claude only showing up in the last month

Maybe you need to chill man

Anonymous wrote:

Oh yeah, "writing"? Yeah, lot's of writing there bro.
[image]
[image]

That image shows a 5 month old repo with Claude only showing up in the last month

Maybe you need to chill man

Literally all of it is Claude https://github.com/jrl290/Reticulum-rust

I am exhausted with being gaslight on my own forum by techbro sloperators.

Literally all of it is Claude

The guy uploaded an empty repo in March? Obviously not

There's no gaslight there. You just think you're providing evidence when you're not.

And before you start calling me more names, I didn't say that he was or wasn't using AI

Anonymous wrote:

Literally all of it is Claude

The guy uploaded an empty repo in March? Obviously not

There's no gaslight there. You just think you're providing evidence when you're not.

And before you start calling me more names, I didn't say that he was or wasn't using AI

It's all Claude. No human person is sitting there writing anything. Show me what button you press on the keyboard for — (U+2014). There isn't one.

This repo claims to also have "LXMF support". It has fields that are made up, and a LXMStamper that isn't compatible with the Python version at all. I do not have time to dig through this shit spewed out by an LLM to tell you if it's "quality". The burden of proof lies on you, the "developer", to do that. People can't even be assed to run a sanity check of whether or not it works with the basic Python provided examples or just works in general. They even outsource that to the LLM.

Chiming in as a fly on the wall here,

I use LLM in development, so by no means am I an LLM hater. The issue isn't LLM, it's slop being shared and unveiled as if it's some showcase item before it's even reached a stable release, and we all know, tools are up to how you use them. And yes, so many things being unveiled aren't even being checked, like someone hammering a nail into a wall, but never looking at it after to make sure the nail is even in. That said, this fight against LLMs here is so tiresome to the point where half the reason I come to this forum anymore is for drama and seeing if Mark has posted for testing. Otherwise this place isn't too useful, and even speaking of Matrix during the switch was also met with your unfriendliness and harshness.

But alas, the fight should be against idiots, not LLMs, as you yourself have clearly used to assist in the making of this forum (granted, we are all just making accusations here based on our analysis of how to catch LLM). And I say this is someone who agrees with most of your takes. I don't support jrl or the frosty guy at all, this is all a time waste. But accusations will always be accusation, however, idiots are idiots and slop is slop, so I prefer to call things for what they are otherwise this battle will never end.

That said, I'm mostly here to defend the emdash. Double tapping the hyphen on most Mac/iOS builds creates a U+2014 emdash automatically. I type 140wpm+ and am generally overly verbose, so em dashes have always been useful for my style.

Granted, we all know the author was using LLMs (as do most developers), but perhaps it's an important factoid to keep in mind for future scenarios. Not all em dashes are evil, AI made them evil, and some systems do allow for it to be easily incorporated — not just with U+2014.

Anonymous wrote:

... and we all know, tools are up to how you use them

It's just a tool, bro

Let me know when a hammer makes the chair you're sitting on all by itself without any oversight besides someone clicking a merge button.

Anonymous wrote:

Let me know when a hammer makes the chair you're sitting on all by itself without any oversight besides someone clicking a merge button.

I think a more apt analogy would be a hammer that you control with your voice. Perhaps useful for tedious tasks in a large project, like hammering in a row of nails while you build a house, but using it to design and build an entire house all on it's own just ain't gonna work, and could very well have catastrophic results.

Claude isn't deciding on it's own that it should write another port of Reticulum, someone is telling it what to do. Until that changes, LLMs will still just be tools. Perhaps very powerful tools, but tools none the less.

cobalt32 wrote:

Anonymous wrote:

Let me know when a hammer makes the chair you're sitting on all by itself without any oversight besides someone clicking a merge button.

I think a more apt analogy would be a hammer that you control with your voice. Perhaps useful for tedious tasks in a large project, like hammering in a row of nails while you build a house, but using it to design and build an entire house all on it's own just ain't gonna work, and could very well have catastrophic results.

Claude isn't deciding on it's own that it should write another port of Reticulum, someone is telling it what to do. Until that changes, LLMs will still just be tools. Perhaps very powerful tools, but tools none the less.

When I've looked at repos, everything from the initial commit to the testing is all written by LLMs. I'm sick of seeing single release repos, all done by LLMs, no support, no updates, nothing. Shit just getting thrown over the fence. There's no care. You have faith that people are using these as a tool and nothing more, I do not see it that way. To continue the analogy, everything from the foundation, scaffolding, all the way to the city inspection is all being done by one non human entity. I would have more faith that code was properly being reviewed, but the commit histories show me otherwise.

Anonymous wrote:

You have faith that people are using these as a tool and nothing more

I do not. I know that people treat LLMs as if they're more than just a tool.

Anonymous wrote:

To continue the analogy, everything from the foundation, scaffolding, all the way to the city inspection is all being done by one non human entity.

This is exactly what I had in mind. Using LLMs in this manner is dangerous. I'm pretty sure we are in agreement.

cobalt32 wrote:

This is exactly what I had in mind. Using LLMs in this manner is dangerous. I'm pretty sure we are in agreement.

I do believe that we are. The rules that have been laid out seem like a positive step in trying to keep things sane here. Every discussion boils down to "it's just a tool", but there is a vast difference when a tool is in the hand of an artisan versus a layman. I hope people can abide by the simple rules provided by the operator