An RTB server for programmatic advertising. RTB — Real-Time Bidding — is when ad exchanges auction ad impressions in milliseconds, every single time a user loads a webpage. Your browser sends a request, and dozens of servers compete to place their ad. All in the time it takes the page to render.
The zero-code challenge
This was my first project where I wrote zero lines of code myself. Literally none. Only prompting and architecture design. I described the components, the data flows, the edge cases — and AI generated the code and tests in Go.
The backstory
A colleague had been building the same kind of server manually for months. I looked at the scope and thought: can I build this from absolute scratch in 3 days (Friday through Sunday), using only prompts? Not a toy — a working PoC with proper tests, budget tracking, campaign evaluation.
The result
The server evaluates which ad campaign to show for each incoming bid request, calculates budget spending in real time, and responds in under 5 milliseconds at 10,000 requests per second. Microservice architecture: Go services communicating via NATS, with BBolt DB for persistent storage and Redis for caching and fast counters.
What I learned
Architecture matters more than code. When you can't write code directly, you're forced to think clearly about every boundary, every data flow, every failure mode. The 3-day constraint also forced brutal prioritization — no gold-plating, no premature optimization, just the core logic that makes money. The tests pass, the latency holds, the budgets are tracked correctly.