Pick a depth. Each prompt opens in your AI pre-loaded with the lesson. Click a row to preview the prompt.
CPU clock speeds plateaued around 2005 when the power wall made further frequency scaling physically impractical — Dennard scaling broke down, and since then silicon has gotten cheaper by adding cores, not cycles. Every modern server has 64–192 hardware threads; a single-threaded program uses less than 1% of available compute on that hardware. Beyond raw CPU, most programs spend the majority of their time waiting — on the network, disk, database, or user input. Concurrency lets you do useful work for request B while request A is blocked waiting for the database, which is why a 200-req/s server with 10ms I/O latency needs thousands of in-flight requests to saturate even a single CPU core.
A modern laptop has 8–12 cores. A server has 64–192. If your program runs on one thread, you're using <10% of the CPU.
Also: most real workloads spend their time waiting — on disk, network, user input. A concurrent program can do useful work for request B while request A waits on the database.
nproc on Linux, sysctl -n hw.logicalcpu on macOS). Calculate how many cores a single-threaded program leaves idle.