Skip to content
AI ConnectPowered by VELENTIS
AI-generated2 min

Autonomous Coding Agents: Simon Willison Demonstrates 3D Browser Game with Audio Multiplayer

Simon Willison built the 3D browser game Battle Drones using Claude Code across 321 commits. The RTS synchronizes multiplayer state via audio signals, highlighting the power of autonomous agents.

This article was AI-generated and published automatically. Context, labelling and all sources at the end of the article.

(KI-generiertes Symbolbild: Gemini / AI Connect)

Software developer Simon Willison has released Battle Drones, a fully playable three-dimensional browser game inspired by the classic real-time strategy title Command & Conquer. The project was constructed without traditional manual programming routines, relying instead almost entirely on autonomous coding agents. Within the international developer community, the experiment is attracting considerable attention as a concrete demonstration of the maturity of vibe coding. Rather than writing code line by line, Willison steered the entire implementation through high-level instructions, deliberately delegating the actual programmatic design and syntax to generative models.

The development chronicle illustrates how far automated software engineering has advanced in practical, production-like environments. The entire project took shape across twelve primary development iterations and comprised a total of 321 individual commits in version control. Willison used the agentic programming tool Claude Code as his primary driver to generate game logic, unit mechanics and visual graphics. Throughout the entire process, he intentionally refrained from inspecting the growing codebase in detail or manually reviewing and fixing underlying functions.

From an architectural perspective, Battle Drones incorporates an unusual and technically challenging multiplayer networking mechanism. Instead of deploying centralized server infrastructure or conventional browser websockets, the system runs a serverless transmission protocol directly inside the web client. State synchronization between two physically adjacent devices is handled entirely through acoustic signals. The participating hardware machines rely solely on their built-in speakers and microphones to transmit, capture and decode game state data locally.

In addition to audio data transmission, the game architecture features battles driven by large language model agents executing tactical decisions in real time. The successful integration of three-dimensional rendering, dynamic game mechanics and local acoustic networking serves as a compelling proof of concept among software engineers. It demonstrates that modern coding agents have moved far beyond generating isolated utility scripts and can independently implement complex, non-trivial system designs. The project proves that even unconventional communication channels and protocols can be orchestrated without granular human code maintenance.

This practical demonstration arrives alongside an intense industry debate regarding the structural evolution of the software engineering profession. Sparked by Jeremy Scheff's widely discussed essay titled 'My obituary for programming', discussions across technical communities like Hacker News have centered on the psychological and operational changes taking place. Historically, software engineers were compensated for solving intricate logical puzzles and handcrafting reliable algorithms. That classic foundation of programming is rapidly eroding as generative systems take over routine implementation and complex programming tasks alike.

According to Scheff's analysis, the core responsibility of developers is quickly pivoting toward overarching system architecture and the active supervision of autonomous refactoring agents. These specialized tools can already optimize, restructure and accelerate existing codebases by a factor of two to three at the touch of a button. For technical professionals, this represents a fundamental transition away from manual line-by-line coding to the role of an architectural director. Projects like Battle Drones demonstrate that this structural redistribution of engineering labor has already transitioned from speculative theory into working software.

What this means for you

For software engineers and technical leads, core competence is shifting rapidly from manual syntax writing to high-level architectural oversight. Professionals who learn to direct autonomous coding tools like Claude Code can capture significant productivity gains, but they must master the skill of establishing rigorous boundaries for agent-generated codebases.

Perspectives

Coverage: 3× Other

One story, several angles: how each source frames the topic, each with a verbatim quote.

  • daily.devOther

    daily.dev highlights how Claude Code enabled a fully functional browser game with audio-based multiplayer and deterministic simulation through vibe coding and targeted prompts.

    Original quote

    „Two devices within microphone range synchronize game state using an audio modem system transmitted as sound signals“

    daily.dev
  • explainx.aiOther

    explainx.ai analyzes the project as a practical case study for builders, detailing the iterative development process, technical workarounds, and the system's real-world limitations.

    Original quote

    „It is a useful case study for anyone building with agents“

    explainx.ai

Source classification is maintained editorially (political spectrum only where consensus is broad; vendor communication is PR, not journalism). Unlabelled sources are unclassified: we do not guess.

Evidence

Solidly sourced
69/100
  • Simon Willison developed the 3D browser game Battle Drones across 12 main iterations and 321 commits primarily using Claude Code, without manually inspecting the codebase in detail.

    verified
  • The game synchronizes multiplayer data serverless via audio signals using the speakers and microphones of two adjacent devices.

    single source
  • In his essay, Jeremy Scheff argues that software engineering is shifting toward system architecture and supervising autonomous refactoring agents that accelerate codebases by a factor of 2 to 3.

    single source

The evidence score is computed, not hand-set: from confidence, the number of sources and the share of verified statements.

Source & transparency

As of: October 20, 2026

AI-generatedAI-generated: produced automatically from vetted sources with technical quality checks (source, quote and figure verification); no human sign-off of each item before publication

Sources
3
Verified statements
1 / 3
Evidence score
69Solidly sourced

Want to put this into practice?

We connect you with suitable AI providers from the DACH region, free of charge and without obligation.

What's next?