Core War

// About this game

Core War is best understood as bot arena built around code as the main verb, not as a normal game with a small programming minigame attached. Core War is the primordial programming game. You write a "warrior" in an assembly-like language called Redcode, and it battles another warrior for control of a circular virtual memory, each trying to make the other execute an invalid instruction. Forty years on, the community still tunes imps, scanners and bombers on shared hills. The useful question for a new player is not simply "is it about programming?", but what kind of thinking it asks for: Assembly, redcode, battle and classic, and a willingness to test an idea by letting the simulation run. Released in 1984 by D. G. Jones & A. K. Dewdney, it sits in the catalog because the program you write is the thing that actually changes the game state.

The core play is adversarial automation. You submit logic, watch it collide with someone else’s logic, study the replay, then decide whether the weakness was sensing, positioning, targeting, risk management or plain overfitting. Good arena games make every loss useful because the opponent exposes assumptions your tests did not cover. In Core War, that means the fun is in the gap between an intention and a working implementation. You start with a rough plan, translate it into the tools the game provides, then watch the result expose every missing condition. A direct solution may pass the first level or match, but the better solutions usually come from noticing a pattern: repeated movement, wasted work, poor targeting, bad routing, a race condition, a blocked path, or a decision that should have been stored as state instead of hard-coded.

The language side is centered on Assembly, but the transferable skill is broader than syntax. You practice decomposition, debugging, iteration and the habit of reading the rules before blaming the machine. The expert rating is earned: this is the kind of game where reading the rules closely, planning on paper and accepting low-level constraints are part of the fun. Because it is offline, it works well as a focused engineering toy: you can pause, restart, inspect mistakes and iterate without the pressure of a live server or a disappearing opportunity. The best sessions are usually not the ones where everything works immediately; they are the ones where a failed run gives you a clear hypothesis for the next version. If the game has leaderboards, ratings or community solutions, those become useful mirrors rather than just bragging rights, because they show how many different shapes a correct program can take.

Core War is strongest for players who like the feeling of making a system slightly smarter each time they touch it. It will be less satisfying if you want fast reflex challenges, cinematic spectacle or a puzzle with only one intended answer. The reward is more specific: seeing your own instructions harvest, fight, route, query, build, solve or survive without your hand on the controls. It is also easy to recommend as a trial because the entry cost is low: you can open it, test whether the programming model clicks, and only then decide how deep you want to go. Taken on its own terms, it is a practical way to turn programming concepts into a visible loop, where every bug is part of the play and every improvement has a concrete effect on the world in front of you.

// What you’ll write

A minimal Redcode warrior copies its current instruction into the next memory cell. Core War strategy starts from tiny patterns like this.

;redcode
;name Imp
MOV 0, 1
END

Core War resources ↗