AI Memory That Actually Works

Phil from Alt.Net shows you the memory patterns that let his AI coding sessions run for days without losing context, live demos included.

Phil Laureano has solved a real problem that anyone using AI for coding runs into: losing context mid-session when the window fills up. He rebuilt an open-source coding harness with a persistent memory system using familiar .NET patterns like pub/sub and topic-based messaging, letting his AI sessions run for days without losing the thread. The talk walks through how friction points drove each architectural decision, from Obsidian markdown files through to the full system. Expect live demos on two terminals showing how agents share knowledge across sessions instantly and how the memory layer turns rough ideas into finished work. This is a working solution you can actually study and adapt, not a theoretical concept.