initing context for blog:

• Currently working on a long term memory system that's less underengineered than those I've seen so far.
• Have a universal board game ai project. Was previously stumped when I realized arbitrarily nested zones didn't sound like they'd fit into ml, but I now think a graph nn might be able to do it and want to see if it's good enough.
• Pray is an immersive sim ttrpg I get hyped about then get stumped on some issue and stop without really making progress
• tabletop teamfight is a tactics boardgame so I can see how abilities from different games would interact.
• I'll eventually get back to trying to get llms good at social deduction games