Make money doing the work you believe in
You have 500,000 pending events spread across time horizons from 1 microsecond to 60 seconds. New events arrive constantly. Others get cancelled. Time advances monotonically, and every event that's due must fire.
Your job: build the fastest event scheduler you can, measured in TSC cycles per operation on bare metal. The catch? We're scoring on P99 latency, not average. Your scheduler needs to be fast on every call — not just most of them. One expensive advance() that fires a thousand callbacks will tank your P99 even if everything else is blazing fast.
The naive solution uses std::multimap. It scores around 6,000 cycles/op at P99. The theoretical floor for a well-designed data structure is orders of magnitude lower.
This is the kind of infrastructure that sits at the heart of every trading system, network stack, and game engine. Getting it right means your order timeouts, heartbeats, and retransmission timers never stall the hot path.
The competition is ongoing. The top 50 get their name on the leaderboard. Our two existing challenges (Order Book and Multi-Symbol Order Book) are still open too — the leader on Challenge 01 is at an absurd 23 cycles/op and showing no signs of slowing down.
