In this episode of The CTO Podcast, Lucas and Luna dive into how Discord built and scaled its real-time messaging infrastructure. Starting with a single Go service handling millions of concurrent WebSocket connections, they walk through the architectural decisions that allowed Discord to support 140 million monthly active users with sub-100ms latency. They cover the shift from a monolithic API to a service-oriented architecture, the use of Redis and Cassandra for state management, and the trade-offs between consistency and availability in a chat system. Luna questions the cost of running a custom game voice network on Erlang, and Lucas defends it with specific numbers on latency and server count. They also discuss what CTOs can learn from Discord's philosophy of optimizing for the 99.9th percentile user, and why sometimes the right technical bet is to ignore the hype and use boring technology.