I would absolutely encourage everyone reading this to check out Reticulum [1] if you haven't already. I believe the base project might be in need of new maintainers(?) at the moment and the main dev has some very strong takes, but it is a very well-thought out approach to distributed networking at the protocol layer. The existing implementations out there include a desktop app which can function over the internet (IP) or a USB connection to some existing LoRA boards. I recently purchased a LilyGo T-Echo [2] and have had a great experience flashing the open-source firmware and using it connected to a desktop over USB or connected over Bluetooth to the fantastic new companion app Columba [3]. This app seriously makes Reticulum feel like a first class citizen when it comes to parity support for messaging. You can even send files/images (with limitations of course)! And since it works at the network level, you can make your own apps to run over Reticulum as well.
I spent an entire month trying to build something with Reticulum, but there just isn’t great tooling for dealing with the protocol. Makes for a pretty infuriating devex if you’re just trying to build your app.
Neat concept but so many footguns that (imo) it’s not really sustainable to try bootstrapping.
Specifically, I had tried to port the stack to Rust no-std to use on nrf52 LoRA devices to use/abuse the existing MeshCore network to deliver reticulum packets. Turned out to be a nightmare just trying to figure out if my packets were even correctly formed.
It's so much fun with little pages, message boards and random people hitting you up for a chat.
I brought up my own transport node and propagation node too to contribute to the mesh.
>We have always been wary of AI generated code, but felt everyone is free to do what they want and experiment, etc. But, one of our own, Andy Kirby, decided to branch out and extensively use Claude Code, and has decided to aggressively take over all of the components of the MeshCore ecosystem: standalone devices, mobile app, web flasher and web config tools.
>And, he’s kept that small detail a secret - that it’s all majority vibe coded.
Without any more context, I am highly suspicious of this framing.
1) Someone "taking over" the ecosystem seems like an entirely different issue. How is this possible? Does it mean he's publishing things and people want to use them?
2) Is the code bad? It sounds like they had no idea he was using AI. That seems to imply there was nothing wrong with the code as-is. Why not judge it on it's merits?
>The team didn’t feel it was our place to protest, until we recently discovered that Andy applied for the MeshCore Trademark (on the 29th March, according to filings) and didn’t tell any of us.
Taking this at face value, this is indeed hostile and bad.
But no, I'm not going to get outraged that someone is simply using Claude Code.
Agreed. I use meshcore and have multiple repeaters setup. I don't care about people using ai assisted coding but I think it should be disclosed especially if its closed source.
Now the trademark take over seems crazy especially given Andy hasn't contributed to the github project, only personal for profit add ons.
I do also think that the meshcore core team have "tacked on" and tried to enforce a stronger narrative with their anti ai coding bias.
> Is the code bad? It sounds like they had no idea he was using AI. That seems to imply there was nothing wrong with the code as-is. Why not judge it on it's merits?
Anyone that has used AI at all knows this isn't how it works. AI is extremely good at producing plausible-but-wrong outputs. It's literally optimised for plausibility, which happens to coincide with correctness a lot of the time. When it doesn't you get code that seems good and is therefore very difficult to judge on its merits.
With human written code it's a lot easier to tell if it's good or not.
There are exceptions to this - usually if you have some kind of oracle like that security work that used AddressSanitizer to verify security bugs, or if you're cloning a project you can easily compare the behaviour to the original project. Most of the time you don't have that luxury though.
What is it with mesh projects and having these super draconian trademark enforcers? Meshtastic is the same. One of the main reasons I got interested in MeshCore was reading the Meshtastic trademark rules and just finding them... really really over the top.
Heh, I use MeshCore in Massachusetts and my layperson explanation is that MeshCore is for people who would be HAMs except they don't have the patience to take an exam.
You're probably more correct, but not having the FCC as a barrier to entry using $20 hardware means a passing curiosity becomes me installing a repeater on our roof with a cavity filter that reaches half a city. It's super fun.
I was using a vibe coded UI (unrelated to this guy) that wasn't super disclosed and each dot revision a new basic thing broke. One I couldn't upgrade the firmware without a full reflash. Now I have to turn bluetooth off and back on to connect to it each time. In both cases it worked fine before that revision came out.
Was it because of vibe coding? I mean... it sure seems likely. Maybe it just needs actual testing?
At the same time it is seemingly the only UI firmware that supports bluetooth to my phone, uses map tiles on an SD card to show GPS maps (I have a tdeck so it has an LCD suitable for it), and runs on a tdeck. Oh, and our local channel names are too long for the ripple firmware (perhaps fixed by now) and the channel number limit was like 4? Maybe 10? Arbitrarily low in any case.
So like... I'm still using this vibe coded UI that breaks some new basic functionality each revision. I can connect to it over bluetooth (even if it's now unreliable), I can use my literally like 1 million map tiles with the GPS, I can actually enter the channel names, and I can have up to 20 channels.
I came up with a way to install a repeater 20ft up a mast that's been on top of the building my office is in, but it's been idle since the TV station that used to be in here left. It has decent reachability, but unfortunately it's not at a particularly high point of the city, it has great reacability into the University and can reach my house, but there's a ridge to the south that puts the antenna more like ground level if it was on that ridge.
Actually the opposite, tons of ppl in the meshtastic community (Discord) berate amateur radio operators. I stopped even discussing the subject because of how much derision I observed or was subjected to. Lots of insults and nasty jokes in passing as soon as the topic even comes up whatsoever. Kinda like your post, actually - offhanded derogatory remarks about an entire group of people solely because of the hobby they're involved in.
IYKYK. Hams are known for a distinctive personality type that can be at strong odds from other tech people and other comms people. Usually in ways that clash with consequences.
I know a few hams that are chill and they are precious doves. I know quite a few more who I won't even engage with for fear of crossing them and them dedicating their lives to making mine hell. Because I've seen them do it to others.
That's not _just_ the hams, mind you. This behavior is overrepresented in hackerspaces in general. But there's a lot of overlap between those groups. Hasn't changed much in the 40-some-odd years I've been involved there either.
I don't really think its fair to lump hams into that behavioral bucket. It's certainly a personality type that tends to get attracted to lots of different technical hobbies.
If you walk around Defcon and keep an eye out for small clusters (2-3) of 40+ yr old white dudes who are loudly, intentionally bitching about something minor/tangential and chasing anyone remotely interesting away from the area by their mere presence..... 100 times out of 10 they are hams.
The phone phreaks have their own quirks too, but it's usually just being smelly and/or chomos. But also that's a tiny percentage. They're way more chill.
Lockpicking is the best village. Honorable mentions to car/hardware hacking and biohacking villages.
For now, I don't think it's fair to compare MeshCore with MeshTastic in terms of enforcement as that has not happened with MeshCore. This seems to be one guy getting a trademark in the UK without the approval of other members of the team. They're not going after anyone. Not yet, at least.
> What is it with mesh projects and having these super draconian trademark enforcers?
Simple. Follow the money. Meshcore has more than 100k of users, repeaters are cropping up like weeds across the world. And that means there is a serious incentive to "cash out".
Notably, the person "cashing out" here wasn't involved in Meshcore firmware or app development, but in marketing.
All meshtastic code is GPL, the name "meshtastic" is owned by the company that developed it. You can use any of the code, you can't use their name outside their rules. This is absolutely no different than, say, Firefox. The trademark policy is very permissable and you don't even need their permission to use the name on a commercial product.
I think it's totally sensible for the organization to want to have some level of control over what gets their label on it -- the Wi-Fi people wouldn't be very happy about someone slapping their logo all over a bunch of completely incompatible hardware.
I've played with MeshCore and Meshtastic a bit, and while they are fun, the general hype seems overblown. The "SHTF" types that get involved with this tend to just taint the whole concept for me. I was/am interested in the use cases for building sensor networks, but most of the chatter seems to be around people who just want to send Hello World type texts back and forth, without realizing how poorly a network like this would perform in a real SHTF scenario.
I largely agree and want to add more,I also think the lack of standards also will effect it's usability in a real shtf scenario. why should I use meshstastic over meshcore for example. I also don't think lora will be in my mind in that kind of scenario.
I feel the same way, and both mobile apps are pretty janky, with Meshtastic being extra obnoxious because the UI teams between Android and Apple apparently don't talk to each other- very hard to onboard/answer questions from someone new if you're on a different platform than them.
It was fun and cheap to set up, but I look forward to something with better messaging persistence so you can at least reliably not miss stuff.
I got to participate in a game that used Meshtastic and GPS where you walk around a large camp and "capture" different regions. It worked great for that and was a lot of fun.
If there ever where a more serious situation where my life depended on one of these meshes, I would be feeling pretty uneasy. They are absolutely not reliable enough to even consider such a thing. I suppose they might be better than nothing.
To say nothing of what is required to set up the devices. I wanted to put a full dev system on a raspberry pi 3 just so it would all be in one place and I could work on it when in a location with no internet - it ran out of memory trying to compile the massive web app that is the default client interface.
Is this client app still closed source? Non-starter for me, also a strong indicator that anything like this was bound to happen, and this will not be the end of it.
Shame to hear that: the protocol works well, scaling up to thousands of nodes across hundreds of miles. This is the local mesh where I live:
https://cascadiamesh.org/map/
You don't have to use the closed source app; there's an open-source client too, there are Blackberry-style client devices which don't require an app at all, and all the actual firmware is open source (MIT).
Wow, the coverage is nuts. I should hop on, looks like I've got solid coverage in my area. Been too lazy to properly give it a try but obviously I really should! Thanks for the link!
It's ridiculous to me that they're concerned about the trustworthiness of AI-generated code when their code quality is so low. They don't even have automated tests and ignore attempts to add them.[0, 1, 2, 3]
Last I checked, there's little validity checking in the code, so it's possible to broadcast nonsense values (like GPS coordinates outside of Earth's bounds) and the code happily accepts it.
And that's fine if they're just like a scrappy upstart doing their best, but it annoys me to be so high and mighty about their code quality when they don't invest in it.
I really want to like MeshCore but I feel like its stewardship makes it hard. The main two people I know running it are Scott Powell and Liam Cottle, both of whom are trying to build businesses on closed-source layers on top of the firmware. I don't think there's anything wrong with an open-core business model (I ran such a business myself), but it creates perverse incentives where the core maintainers try to suppress information about the open-source alternatives and push their own closed-source paid products.
Also, MeshCore's recommended broadcast settings for the US are illegal.[4] I emailed the Liam and Scott about this months ago, and they ignored me.
This is also a loaded question. The only specifics they've offered are that he simply used Claude Code. Um... OK? Do the tests pass? Did his changes add any security flaws? Regressions that were untested?
What do you mean? Is the non-existent millionth floor of the Empire State Building still part of the Empire State Building?
Also, I'm assuming we're in agreement that software should not accept invalid GPS coordinates from untrusted peers regardless of semantics about whether or not they're within Earth's bounds.
> It's ridiculous to me that they're concerned about the trustworthiness of AI-generated code when their code quality is so low.
Agreed, but at least it's somewhat sensibly structured. AI? Good lord you'll end up with a slopaghetti mess.
> They don't even have automated tests and ignore attempts to add them.[0, 1, 2, 3]
Two people, 540 issues and 270 PRs open at the moment. Not wanting to be that guy... but do the math. The reviewer team is small as hell and after this drama (which probably kept both of them busy with BS) they'll likely be even less willing to trust others.
If you want to stand a better chance at getting your code into other people's hands, go and contact the person behind the Evo fork. IIRC he's part of Hansemesh, Germany's biggest regional MC.
I have heard indirectly multiple times now that the only two ways to get a PR of interest merged is to either gather enough people to Like the issue on Github or to join the Discord and ask.
Spicy take but I think this whole Lora thing is a dead end bad technology not worth our time.
Wifi HaLow 802.11ah is finally out & available, sometimes at ok prices. We don't really have ad-hoc communication pioneered for wifi, but it's doable and we ought lean into it, rather than using some totally different stack, especially one that is under strict control of a single company.
What we learn doing wifi halow can directly port and improve the rest of ways we connect. That would be grand.
Eh, I've played with 802.11ah and it's very long range for the power, but way too stateful for the use case that these mesh protocols serve. If you go out of range it's several seconds to re-establish a connection. I've worked with proper mesh radios (Silvus, etc) and they work very differently.
I love using AI to develop and I think it is important in modern development, but you definitely have to disclose it because there is a difference between AI and human written code is key.
You're right, I don't know who would down vote or disagree with this. It's not easy to figure out what a program is doing, but if someone wrote it you at least know there was some sort of intention there. With AI you have no idea why it's there.
All over the internet people are putting up vibe coded projects and no one says that's what it is up front. They all just say "I made this" and they are more than happy to take in the adulation of people impressed that they made something with an animated pattern.
Then when they finally admit that they wrote nothing and don't know how any of it works people start to say "nothing wrong with using ai", as if using it is the same as copying it verbatim, not understanding anything and taking complete credit as if you wrote it while lying about how it was used.
Ah, the cliche story of most open source projects! Still, that was a dick move to hijack the project just because you were the “front end” or the promoter, and found that AI can help so you backstabbed the team and hijacked the trademark as well, class A a-hole!
Just a word of caution, claude code might look impressive if your knowledge is shallow or intermediate in a topic, but if you know what you’re doing, once you dig deeper it starts to introduce plenty of scope creeps into your code base, piling one on top of another that you won’t even recognize your own code shortly.
[1] https://reticulum.network/ [2] https://lilygo.cc/products/t-echo-lilygo [3] https://github.com/torlando-tech/columba
Neat concept but so many footguns that (imo) it’s not really sustainable to try bootstrapping.
Specifically, I had tried to port the stack to Rust no-std to use on nrf52 LoRA devices to use/abuse the existing MeshCore network to deliver reticulum packets. Turned out to be a nightmare just trying to figure out if my packets were even correctly formed.
Different countries allow unlicensed use on different frequencies. Look up which is correct for your location.
Only very very small testbeds.
https://rmap.world/
It's so much fun with little pages, message boards and random people hitting you up for a chat. I brought up my own transport node and propagation node too to contribute to the mesh.
at the very least, try it. maybe it's simpler than you think
>And, he’s kept that small detail a secret - that it’s all majority vibe coded.
Without any more context, I am highly suspicious of this framing.
1) Someone "taking over" the ecosystem seems like an entirely different issue. How is this possible? Does it mean he's publishing things and people want to use them?
2) Is the code bad? It sounds like they had no idea he was using AI. That seems to imply there was nothing wrong with the code as-is. Why not judge it on it's merits?
>The team didn’t feel it was our place to protest, until we recently discovered that Andy applied for the MeshCore Trademark (on the 29th March, according to filings) and didn’t tell any of us.
Taking this at face value, this is indeed hostile and bad.
But no, I'm not going to get outraged that someone is simply using Claude Code.
Now the trademark take over seems crazy especially given Andy hasn't contributed to the github project, only personal for profit add ons.
I do also think that the meshcore core team have "tacked on" and tried to enforce a stronger narrative with their anti ai coding bias.
In that context it is quite logical to take a trademark out once the project is mature enough so you can profit off other people's work.
Considering their user base does not like the hidden vibe coded idea I don't think this is bias but a sane rationalisation.
Anyone that has used AI at all knows this isn't how it works. AI is extremely good at producing plausible-but-wrong outputs. It's literally optimised for plausibility, which happens to coincide with correctness a lot of the time. When it doesn't you get code that seems good and is therefore very difficult to judge on its merits.
With human written code it's a lot easier to tell if it's good or not.
There are exceptions to this - usually if you have some kind of oracle like that security work that used AddressSanitizer to verify security bugs, or if you're cloning a project you can easily compare the behaviour to the original project. Most of the time you don't have that luxury though.
You're probably more correct, but not having the FCC as a barrier to entry using $20 hardware means a passing curiosity becomes me installing a repeater on our roof with a cavity filter that reaches half a city. It's super fun.
I was using a vibe coded UI (unrelated to this guy) that wasn't super disclosed and each dot revision a new basic thing broke. One I couldn't upgrade the firmware without a full reflash. Now I have to turn bluetooth off and back on to connect to it each time. In both cases it worked fine before that revision came out.
Was it because of vibe coding? I mean... it sure seems likely. Maybe it just needs actual testing?
At the same time it is seemingly the only UI firmware that supports bluetooth to my phone, uses map tiles on an SD card to show GPS maps (I have a tdeck so it has an LCD suitable for it), and runs on a tdeck. Oh, and our local channel names are too long for the ripple firmware (perhaps fixed by now) and the channel number limit was like 4? Maybe 10? Arbitrarily low in any case.
So like... I'm still using this vibe coded UI that breaks some new basic functionality each revision. I can connect to it over bluetooth (even if it's now unreliable), I can use my literally like 1 million map tiles with the GPS, I can actually enter the channel names, and I can have up to 20 channels.
I know a few hams that are chill and they are precious doves. I know quite a few more who I won't even engage with for fear of crossing them and them dedicating their lives to making mine hell. Because I've seen them do it to others.
That's not _just_ the hams, mind you. This behavior is overrepresented in hackerspaces in general. But there's a lot of overlap between those groups. Hasn't changed much in the 40-some-odd years I've been involved there either.
I always found it interesting how many useful little apps hams write, keep them closed source and then...die.
The phone phreaks have their own quirks too, but it's usually just being smelly and/or chomos. But also that's a tiny percentage. They're way more chill.
Lockpicking is the best village. Honorable mentions to car/hardware hacking and biohacking villages.
Simple. Follow the money. Meshcore has more than 100k of users, repeaters are cropping up like weeds across the world. And that means there is a serious incentive to "cash out".
Notably, the person "cashing out" here wasn't involved in Meshcore firmware or app development, but in marketing.
I think it's totally sensible for the organization to want to have some level of control over what gets their label on it -- the Wi-Fi people wouldn't be very happy about someone slapping their logo all over a bunch of completely incompatible hardware.
It was fun and cheap to set up, but I look forward to something with better messaging persistence so you can at least reliably not miss stuff.
If there ever where a more serious situation where my life depended on one of these meshes, I would be feeling pretty uneasy. They are absolutely not reliable enough to even consider such a thing. I suppose they might be better than nothing.
To say nothing of what is required to set up the devices. I wanted to put a full dev system on a raspberry pi 3 just so it would all be in one place and I could work on it when in a location with no internet - it ran out of memory trying to compile the massive web app that is the default client interface.
The closed client isn't needed anymore.
My local mesh was testing out meshcore last week, this definitely kills my interest too
https://github.com/zjs81/meshcore-open
You don't have to use the closed source app; there's an open-source client too, there are Blackberry-style client devices which don't require an app at all, and all the actual firmware is open source (MIT).
Worth noting that the Blackberry-style devices are also closed source and the hardware and software is way worse than Blackberry was 22 years ago.[0]
[0] https://mtlynch.io/first-impressions-of-meshcore/#this-is-no...
I wasn't expecting the T-Deck to be anything more sophisticated than a walkie-talkie for SMS, but it's a bummer than the UI code isn't open.
[1]: https://meshcore.io
[2]: https://meshcore.co.uk
It's ridiculous to me that they're concerned about the trustworthiness of AI-generated code when their code quality is so low. They don't even have automated tests and ignore attempts to add them.[0, 1, 2, 3]
Last I checked, there's little validity checking in the code, so it's possible to broadcast nonsense values (like GPS coordinates outside of Earth's bounds) and the code happily accepts it.
And that's fine if they're just like a scrappy upstart doing their best, but it annoys me to be so high and mighty about their code quality when they don't invest in it.
I really want to like MeshCore but I feel like its stewardship makes it hard. The main two people I know running it are Scott Powell and Liam Cottle, both of whom are trying to build businesses on closed-source layers on top of the firmware. I don't think there's anything wrong with an open-core business model (I ran such a business myself), but it creates perverse incentives where the core maintainers try to suppress information about the open-source alternatives and push their own closed-source paid products.
Also, MeshCore's recommended broadcast settings for the US are illegal.[4] I emailed the Liam and Scott about this months ago, and they ignored me.
[0] https://github.com/meshcore-dev/MeshCore/pull/925
[1] https://github.com/meshcore-dev/MeshCore/issues/1059
[2] https://github.com/meshcore-dev/MeshCore/pull/1065
[3] https://github.com/meshcore-dev/meshcore.js/pull/11
[4] https://github.com/meshcore-dev/MeshCore/issues/945
This is also a loaded question. The only specifics they've offered are that he simply used Claude Code. Um... OK? Do the tests pass? Did his changes add any security flaws? Regressions that were untested?
Also, I'm assuming we're in agreement that software should not accept invalid GPS coordinates from untrusted peers regardless of semantics about whether or not they're within Earth's bounds.
Circular coordinates wrap around, cartesian coordinates don't.
Agreed, but at least it's somewhat sensibly structured. AI? Good lord you'll end up with a slopaghetti mess.
> They don't even have automated tests and ignore attempts to add them.[0, 1, 2, 3]
Two people, 540 issues and 270 PRs open at the moment. Not wanting to be that guy... but do the math. The reviewer team is small as hell and after this drama (which probably kept both of them busy with BS) they'll likely be even less willing to trust others.
If you want to stand a better chance at getting your code into other people's hands, go and contact the person behind the Evo fork. IIRC he's part of Hansemesh, Germany's biggest regional MC.
I have heard indirectly multiple times now that the only two ways to get a PR of interest merged is to either gather enough people to Like the issue on Github or to join the Discord and ask.
[1] https://github.com/mattzzw/MeshCore-Evo
Wifi HaLow 802.11ah is finally out & available, sometimes at ok prices. We don't really have ad-hoc communication pioneered for wifi, but it's doable and we ought lean into it, rather than using some totally different stack, especially one that is under strict control of a single company.
What we learn doing wifi halow can directly port and improve the rest of ways we connect. That would be grand.
It is essential to disclose it.
All over the internet people are putting up vibe coded projects and no one says that's what it is up front. They all just say "I made this" and they are more than happy to take in the adulation of people impressed that they made something with an animated pattern.
Then when they finally admit that they wrote nothing and don't know how any of it works people start to say "nothing wrong with using ai", as if using it is the same as copying it verbatim, not understanding anything and taking complete credit as if you wrote it while lying about how it was used.
Just a word of caution, claude code might look impressive if your knowledge is shallow or intermediate in a topic, but if you know what you’re doing, once you dig deeper it starts to introduce plenty of scope creeps into your code base, piling one on top of another that you won’t even recognize your own code shortly.