Player programming

The player authors programs in a textual or visual language that the game interprets each tick to drive entities. State owned: program source, interpreter state (instruction pointer, registers, memory), and an execution budget. Update rule: each tick, run each program for up to a budget of instructions or CPU time, read sensors through an API, and issue commands that the simulation resolves. Languages range from real ones (JavaScript in Screeps and Bitburner, Python-like code in The Farmer Was Replaced) to assembly-like puzzle languages (TIS-100, Shenzhen I/O, Exapunks) and visual instruction tracks (SpaceChem waldos, Opus Magnum arms). Screeps is canonical for a persistent world: each player's main module runs once per tick against a fixed state, commands are applied together at tick end, and a CPU limit with a carryover bucket constrains computation.

The player designs behavior rather than performing it. Writing, running, and debugging a program that then works unattended is satisfying in a way direct control is not, and failures are logic errors the player can find and fix.

• language: real general-purpose, constrained assembly, or visual
• execution budget per tick (instructions or CPU milliseconds)
• sensor and actuator API surface
• debugging tools: stepping, logs, breakpoints, visualization
• persistence: programs keep running while the player is offline
• scoring of programs by size, speed, or cost

• A steep entry cost excludes players who cannot already program.
• A real language with a rich API lets players import libraries and copied code, so the community's code replaces the puzzle.
• Execution budgets set too tight make debugging feel like fighting the platform.
• Weak debugging tools turn failures into opaque guessing.

• Automation through code lets players operate at scales no one could control by hand.
• In shared worlds, players' programs compete directly, producing an arms race of strategies and exploits.
• Code sharing builds communities around libraries and frameworks.

• When code assistants can write competent bots from a description, what remains of the programming puzzle, and does play shift to specification and review?
• Can games teach programming effectively by letting AI assistants scaffold and then withdraw?