Why Your Coding Agent Forgets Everything

Why Your Coding Agent Forgets Everything

Modern coding agents can write code, run tests, and solve complex tasks, yet most start every session with no memory of what happened before. In this talk, we’ll explore why today’s agents are fundamentally stateless, how memory changes their behaviour, and what happens when agents can learn, remember, and improve over time. Through real-world experiments building large-scale agent systems, we’ll look at the path from simple LLMs to persistent, memory-driven software systems.