Anyone else confused how agent memory actually works?
Ada yang tau... wait, I mean, anyone know how agent memory really works? I've been messing with AI agents for my workflow and the 'memory' part always feels like magic until it forgets something dumb. Lately I've been trying to figure out what's actually stored versus what gets pulled back in each run, and it's way more layered than I expected. Is anyone building agents that actually remember stuff well across sessions, or is it still mostly hacks lol? Curious what setups you guys are using.