All work
Systems Programming

Rapid Prototyping with AI Coding Agents

Used classic, well-understood games as a fast-iteration test bed to explore how far AI coding agents can be pushed when paired with a disciplined, spec-driven development process.

Spec → plan → code → review

Process

3+ (movement, visual overhaul, standalone)

Iterations built

The Problem

AI coding assistants are only as good as the process around them. The goal was to treat the AI agent like a capable pair-programmer working from a written spec and plan, with explicit design iteration, rather than accepting whatever a single prompt produces.

My Approach

  • Built multiple iterations of a Snake game, each driven by its own written design spec and implementation plan before code was generated — a first pass focused purely on movement feel (smooth, sub-pixel-interpolated motion at a high frame rate instead of classic grid-snap movement), a second, later pass was a full visual overhaul with a deliberate retro-terminal aesthetic, planned and specified as its own scoped project.
  • Built a separate, complete standalone implementation (grid-based movement, collision detection, persistent high score, progressive difficulty, full keyboard control scheme, mobile-responsive layout) as a clean, from-scratch reference implementation.
  • Used the process itself as the actual point of the exercise: writing a spec, writing an implementation plan against that spec, generating/reviewing code against the plan, and iterating — a repeatable workflow that generalizes well beyond games to any AI-assisted development task.

Stack

Core

HTML5 CanvasVanilla JavaScript

Persistence

Browser localStorage

Process/Tooling

AI coding agentsWritten design-spec and implementation-plan documents as a development discipline

Skills Demonstrated

  • ▸Front-end game-loop programming fundamentals (movement, collision, rendering, input handling)
  • ▸A mature, repeatable process for working effectively with AI coding agents: spec first, plan second, code third, review always
  • ▸Product thinking even on a "toy" project — treating feel, polish, and visual design as first-class requirements, not afterthoughts