Category: Technology

Stats at reticulum.thejeshgn.com 0

A Public Reticulum Backbone Interface in India

Until last week, India had no Public Backbone Interfaces for the reticulum. I started one last week so we could have one in India. I really wanted one in Bangalore, but the Mumbai server was the cheapest. So I set up one in Mumbai. Of course, you can use the one outside India;...

Sideband is an extensible LXMF messaging and LXST telephony client 0

Messaging with Reticulum and LXMF

The first kind of application you could try after setting up Reticulum, or with it, is messaging. I think Reticulum protocol designer Mark Qvist knew that without a practical example of the protocol’s use, people might not use it. Messaging seems obvious. LXMF (Lightweight Extensible Message Format) is a universal messaging protocol built...

A Reticulum node mapped on Rmap World. 0

Getting Started with Reticulum

I have been looking for a network protocol stack that can be run in addition to the regular TCP/IP network at TVC. Kiran has been exploring Meshtastic, mainly to connect the sensors to the Home Assistant that TVC runs. We have very poor cell phone connection and two BSNL wired internet connections, one...

Forerunner 965 showing recovery 0

Why Garmin Forerunner 965?

I have been looking for a full-fledged smartwatch for a while now. I have a Pebble and a BangleJS; I use them all the time. But they are semi-smartwatches. I wanted a full-fledged watch that lasts for days and works fully offline. This post isn’t a review; that will take time. This one...

Offline Maps on Garmin Forerunner 965. 0

Loading Indian Maps on a Garmin Forerunner 965

The primary reason I bought this watch was the ability to display and navigate using proper offline maps. Garmin doesn’t have maps for India. But the OpenStreetMaps community has maps that can be loaded into Garmin. There are many community-built web services, such as Thukke’s OSM Downloads and Userbeam, that provide Indian maps...

4

Exploring 1-Bit LLMs

My interest in LLMs has increasingly shifted towards smaller models, small in terms of size and their ability to run on modest local machines. For some time now, I have been using models with fewer than 40 B parameters on my machines in some quantized way. Then I came across BitNet, a project...