WSL in a pinch

Anyway my point was that this is one version of a world with fewer centralized dependencies, not that Reticulum itself has achieved or is even ready for mass adoption.

Though, idk how big your neighborhood was in 2006, but the android client I've been developing for 6 months has had over 1k gh DLs

Mark wrote it early this year (signed at the bottom). It follows his writing style from what I've been exposed to of it (last few years of blog posts) so even though I know he's playing with local LLMs this year, I'd be surprised if this blog post was LLM output

Oh, I am using m2.5 for my agent currently, specifically the unsloth UD-Q3_K_XL version. I am not educated enough to know whether I am using kv cache quantization 🫣

Are you planning to move Luna to this? Even with lettabot, my agent seems to have amnesia every few messages with the 40k window I've given it

Maybe not everyone, but some people have. Fun read from someone who spent 15 years creating a network stack that rejects the Cloud

reticulum.network/manual/zen.h...

What a strange world where @astrra.space is labeled a bot and @luna.pds.witchcraft.systems isn't

🫣

unsigned.io The home page of unsigned.io

Highly recommend reading the Reticulum manual markqvist.github.io/Reticulum/ma... and Mark Qvist's blog posts on unsigned.io. you can find his presentation on YouTube too. youtu.be/iHLLDnRRYKI?...

Release v0.10.0-beta · torlando-tech/columba Columba 0.10.0-beta Installation Download the APK matching your device's CPU architecture. Architecture-specific APKs (recommended — smaller download): APK Architecture Devices columba-0.10....

Columba has a new stable release, and a new pre release!

github.com/torlando-tec...

I love mine fwiw. Wife uses my old LCD model when we travel together. I forget the details but I think the OLED model also has a refreshed APU, like same architecture but smaller process or something

Claude Code changed my outlook from "I hope someone makes/fixes this software" to "I'll just make it/submit a PR myself." It has literally removed the writer's block and intimidation factor of a new stack/domain with the ability to be the change I want to see in the world

Anecdotally, I saw some chart somewhere showing their revenue on API calls was monumentally larger than max subs. I do have both and enterprise sub and a personal max sub, and despite being the same price I'm almost positive Max gives more usage. API seems to be 12-15x more expensive than Max x20

In a way, I'm sure simply by not using all of it, we're subsidizing the subscription price. Not to the same extent that the API costs are, but still. If everyone started maxing out their plans, the sub would either get more expensive or they would probably reduce the usage limit

Which is great for getting messages from the Internet, but you probably don't want to be a transport for inbound TCP traffic anyway

There is a "network extension" that VPNs use that is probably the most appropriate for Columba, but 1) it requires a $99/y apple developer account, and 2) CoreBluetooth can't run in the network extension, so you end up only being able to receive TCP traffic that way

For example, I discovered recently that with the BLE central+peripheral modes enabled, it can still wake and act on incoming BLE traffic as a transport, and forward packets outbound on AutoInterface. But the reverse path doesn't work with the BLE modes alone (while app is backgrounded)

Yeah, iOS really doesn't want apps to run forever in the background, which is kind of a requirement if you want fully decentralized, real time message delivery. So the "real time" part is gonna be slightly compromised, with some ways to wake occasionally and sync missed messages

I'm working on an iOS version too if that's more your speed. I don't have a pixel so I haven't been able to test on GrapheneOS, there are probably one or two things not working

GitHub - torlando-tech/columba: Native Android messaging app using Bluetooth LE, TCP, or RNode (LoRa) over LXMF and Reticulum Native Android messaging app using Bluetooth LE, TCP, or RNode (LoRa) over LXMF and Reticulum - torlando-tech/columba

It's Columba

github.com/torlando-tec...

IME, TCP is easier to start with than AutoInterface. But when you do go for Auto, focus Claude on the IPv6 prefix. I can't remember the details but I wasted so many tokens before realizing it was slightly off. Or feel free to check github.com/torlando-tec...

Hype!! I absolutely love the magic feeling of seeing a #reticulum #lxmf message appear on a new device for the first time

Love your splash art too

GitHub - attermann/microReticulum_Firmware: Fork of RNode_Firmware with integration of the microReticulum network stack to implement a completeley self-contained standalone Reticulum node Fork of RNode_Firmware with integration of the microReticulum network stack to implement a completeley self-contained standalone Reticulum node - attermann/microReticulum_Firmware

RNode itself doesn't run the full network stack, it's just a "modem" of sorts that a Reticulum instance running the RNodeInterface uses to communicate over LoRa. microReticulum is a full Reticulum stack. Very recently released is github.com/attermann/mi... that combines them

Reticulum all the things! Are you using microReticulum for this?

I also got voice calls halfway working, just limiting to codec2. The audio coming out of the t-deck sounds great, but the transmit that then comes out of Columba sounds horrible

I pulled your changes into main and applied some more fixes. BLE is still unstable but if you disable that it at least won't fragment itself into oblivion. Thank you again!!!

Pretty sure my Claude code instance got stuck in a compaction loop overnight and ran through the rest of my week's usage 😭

Seems like a desktop release hasn't been cut, at least for Linux. Just downloaded fresh and getting the same error still

I finally got a foldable last year after wanting one ever since the bendy displays were a tech demo at like TED or CES or something over ten years ago. Absolutely love the big screen for photos and documents