Oh, exciting progress on my new processing/analytical "engine". I've built a system to break down complex processing and analytical jobs into more "atomic" tasks. These tasks are defined as simple JSON-data structures and can then be sent to an LLM-engine of choice. For example one thing I often do when reading articles it to try and figure out which entities are the attackers and victims. Then I tag this in order to be able to quickly understand context of a mentioned threat actor etc. This engine allows me to do this automatically with a better than I hoped for accuracy. Entities like countriese, cities, places, people, groups, and actors are all classified as attackers, victims or neutral. This contributes to significant speed-up in my analytical work because I don't have to spend time tagging and otherwise identifying these "obvious" but very time consuming tasks. This is so much fun!