honestly there is something extremely cyberpunk about the idea of being able to message an LLM agent running off solar power over an encrypted local radio mesh network even when the rest of the power grid is down, together with the cell towers and the fiber backbones

RE: https://bsky.app/profile/did:plc:nbfjoeficjzf3pejpontvril/post/3meq2jkzbvc2m
running a local model?
yup
whaja pick? I'm currently evaluating glm-4.7-flash Q4KM on the framework desktop via llama-cpp-vulkan. ~45 t/s of output
minimax m2.5 UD-Q3_XL, ~20tok/s
here's my nix flake config in case it's helpful, beware it's slop. open to suggestions gist.github.com/kumavis/0640...
ai.nix

GitHub Gist: instantly share code, notes, and snippets.

Gist
i do t use nix but the info in the flake is pretty outdated, gfx1151 is supported by rocm now and running llama.cpp in vulkan mode is much much slower, i recommend you look into that just make sure the VRAM in bios is set to 512MB and set the GTT to something like 124GB for shared ram/vram
#DMRG is also planning to build a cluster of local reticulum networks based in Yokohama for its archival project. we are kinda uncertain about the number of LoRA cell towers in the 閒東 area…is Tokyo covered well enough to support a 923MHz reticulum mesh?
unfortunately, to the best of my knowledge, my LoRa node is literally the only reticulum node in tokyo at the moment but we should definitely link ours together, at least over an IP backbone!
linking our nodes together sounds great! let’s share our reticulum info as soon as we construct ours (lilygo T-beam 923MHz in εƒηŸ³ι›»ε•† sounds promising). can DMRG make some form of alliance with witchcraft.systems?
Witchcraft Systems

a network of solar powered webring might be a good joint project
yeah, the t-beam should be good enough to get you started, though i really recommend you get an OLED screen for it (since it doesn't come with one by default, you can prolly ask the staff and they'll help you) ideally you want something like a t-beam supreme from aliexpress though, at a later point
our node is actually located very close to akihabara so you should be able to reach us if you use something like columba and configure your node with the same settings as ours (listed on 303058b1c1ca6d0b8574509877d4d4c2:/page/index.mu, reachable over the IP backbone on nomadnet)
it only announces once an hour as usual, but you can DM e001c16837985f21f5a322efcd7565a2 to test for connectivity, that address is always live and monitored (by @[email protected])
as for the collab - witchcraft.systems is mostly in LTS mode right now due to main members being pretty busy with life, so unfortunately collaborations in that regard will have to wait until a better time...
Witchcraft Systems