A few months ago I set myself an ambitious technical challenge: building a system capable of claiming Discord usernames in under 15 milliseconds. What started as a personal experiment became a platform called blosm, which now handles over 1,200 active users with a 99.9% success rate. In this article I share the architecture details, real challenges, and how experience in custom software development was key to achieving it.
The core problem is that Discord's public API is designed for general use, not speed. When a username becomes available, hundreds of people try to claim it simultaneously. The official client adds overhead through HTTP connections, JSON serialization, and TLS handshake latency on every request. While a normal request takes 50–200 ms, a raw-socket sniper can complete the operation in under 15 ms.
blosm's architecture bypasses Discord's HTTP API entirely. Instead, it communicates directly through WebSocket connections at the socket level. This eliminates repeated connection setup, TLS handshakes (the connection is persistent), and JSON parsing on the hot path. The claim engine is written in C++ with a non-blocking socket pool handling thousands of concurrent connections on a single thread. We use lock-free data structures and memory-mapped username queues for zero-copy operations.
To avoid detection, we implemented a multi-layer anti-detection system: SSL fingerprint rotation (JA3) to mimic different clients, a pool of over 200 realistic user agents, request distribution across 4,200 residential and datacenter proxies, and temporal jitter to imitate human behavior. The proxy manager health-checks every proxy every 30 seconds and automatically removes dead ones. During peak hours, we handle over 50,000 connection attempts.
The auto-snipe engine monitors username availability in real time using user-defined regex patterns. When a name is detected as free, the priority is assigned according to the user's plan, the optimal token and proxy are selected, and the claim fires in under 15 ms. The entire process, from detection to webhook notification, does not exceed 200 ms.
Scaling to over 1,200 concurrent users was not trivial. Each user can run up to 10 active patterns, resulting in over 12,000 simultaneous monitors. To manage Discord's gateway connection limits, we multiplex multiple users over shared connections using session management. Additionally, we maintain a pool of tokens per user that are automatically rotated when rate-limited. The infrastructure relies on Docker containers on bare metal, with PostgreSQL and Redis as databases, and a custom metrics pipeline for monitoring.
Several incidents occurred during development: a connection pool leak in week 2, a Discord gateway protocol change in month 1, and a user launching 500 patterns simultaneously causing a DoS on the claim queue. Each problem forced us to iterate quickly and learn. That adaptability is exactly what we apply in cloud environments like AWS or Azure to ensure high availability and scalability.
Today blosm has an average claim latency of 8–15 ms, 99.9% uptime, and over 4,200 proxies in the pool. But this is just the beginning. Next steps include AI-powered name prediction, multi-platform support (Roblox, Twitch, etc.), a public API for developers, and a browser extension for real-time availability checking. Integrating AI agents will allow us to anticipate which names will become free and optimize resource allocation.
Behind this project are months of work fueled by curiosity, caffeine, and too many Discord API documentation tabs open. If you are interested in the technical side or want to learn more about how we build high-performance systems, we invite you to explore the services of Q2BSTUDIO, where we combine Business Intelligence with Power BI, cybersecurity, and automation to create custom solutions. From native cloud applications to algorithmic trading systems, our approach is always the same: minimal latency, maximum reliability.
