Aurora 4X is a legendary freeware space strategy game, first released in 2004 by its creator, Steve Walmsley, and rebuilt in C# in 2020. It is, by any measure, one of the deepest strategy games ever made — a turn-based 4X sandbox of exploration, industry, and empire-building across a procedurally generated galaxy. Rather than real-time play, you select a time increment — from a few seconds up to thirty days — and the game processes that span as a single turn, then waits for your next command.

What sets Aurora apart is the depth of its systems. Every ship is a custom assembly of components — engines, weapons, sensors, fuel tanks, crew quarters — designed by the player. Civilian economies run autonomously, with shipping lines and mining companies responding to supply and demand. Officers have personalities, skills, and careers. NPC empires pursue their own agendas. The community famously calls it “Dwarf Fortress in space.”

Aurora was also never just about strategy — it was a role-playing experience. With no win condition, it functioned as a story generator, and players shared their sprawling interstellar sagas as After Action Reports. It proved that a dedicated audience exists for genuinely complex, detail-rich space games.

But Aurora was always one person’s passion project — a hobby game built for its creator that he happened to share with the world. It was never designed to be approachable, moddable, or to perform well as empires grew large. Its interface is a dense wall of spreadsheets with no graphics, its learning curve is famously steep, and its architecture is closed to community modification. None of this diminishes what it achieved; it simply points to the opportunity Protostar exists to pursue.

Aurora 4X proved something important: that players exist who want real strategic depth — the kind where you design every component of every ship, manage civilian economies across dozens of star systems, and guide your empire's growth and prosperity across centuries of game time. Its creator, Steve Walmsley, built something genuinely unique, and this project exists because of that inspiration.

Aurora was also never just about strategy. It was a role-playing experience — commanders with personalities and careers, empires with distinct identities, and emergent stories that players shared as sprawling After Action Reports. The game was a story generator as much as a strategy game.

But Aurora was always one person's passion project. It was never designed to be approachable, moddable, or performant at scale. And that's okay — it was never meant to be. The result is a game with unmatched depth but a learning curve so steep that most players never get past the first screen, an interface of dense spreadsheets with no visual feedback, and a completely closed architecture that forbids community modification.

Protostar exists to make Aurora's vision accessible to everyone. Same depth. Same attention to detail. Same emergent stories of interstellar empires. But built on a modern engine that performs in real-time, with a browser-based UI that's approachable and visual, an open modular architecture designed for community extension, built-in tutorials so new players can learn at their own pace, and a focus on onboarding so more people can discover what makes this genre special.

• Real-Time Performance — An ECS-based engine built to handle thousands of entities simultaneously, eliminating the long waits between turns in mid to late game that Aurora players know all too well.
• Approachable UI — A browser-based interface with visual feedback, tooltips, clean layouts, toast notifications, and a quick command palette for fingertip access to frequent commands, so the depth lives in the gameplay, not in navigating the controls.
• Built-In Tutorials — Integrated tutorials that let new players learn at their own pace, lowering the steep learning curve that keeps most people away from this genre.
• Moddable Architecture — A modular design where the core game is itself a module. The community can extend, modify, or replace systems — the game grows with its players. A community hub built right into the game lets you browse, install, and manage mods without ever leaving Protostar.
• Cross-Platform & Multi-Window — Runs anywhere .NET and a browser can reach. Multi-device support lets you open multiple tabs — system view, galaxy map, logs — all connected to the same game.
• Multiplayer — Share a galaxy with friends. Cooperative empires, shared discovery, and the emergent stories that only happen when more than one mind is at the table — a frontier Aurora never explored.

^* Protostar is in active, early development. The points above describe the project's goals and intended direction — the destination we are building toward — rather than a list of features that are all complete today. Capabilities will grow and mature as development advances. See the roadmap for where things stand and what's coming next.

Protostar's moddability isn't bolted on — it comes from how the project is split into three distinct layers, each with a clear job. Understanding them explains why the core game is itself a module, and why mods can change almost anything without fighting the engine.

This separation is what lets the community go as far as they want — from small flavor tweaks and balance changes to total conversions that replace the Core Mod entirely. The engine never gets in the way.

What makes Protostar unusual isn't just what it aims to be — it's how it's being built. The entire codebase is generated and maintained through prompt engineering with open-source LLM models, and the game's art is created with Qwen-Image. By intent, roughly 99% of the project is produced and curated through AI.

It didn't start as a plan. Back in December 2025, this was an unintended curiosity — an experiment with the early days of “vibe coding.” Having heard what was becoming possible, I told myself I'd go all in on the wildest project of my dreams and see how far it would actually go, fully expecting to give up somewhere down the line when the results turned terrible.

That's not what happened. The more I prompted, the more invested I became. As a direction and a set of objectives slowly emerged and took shape, there came a point mid-development where I realized this could in fact become a real project — one I could publish and share with the Aurora community and other 4X fans. From there, the work shifted from experiment to genuine effort: turning it into an actual playable game.

That effort continues today. I intend to keep refining, extending, and enhancing Protostar with the goal of someday reaching every objective set for the project — a modern, accessible, open tribute to Aurora, built in close partnership with AI.

There's a widespread and understandable opinion that “vibe coding” — building software by prompting AI — produces bad, sloppy, insecure code. I respect that view, and I understand where it comes from. A lot of AI-assisted code out there genuinely deserves the criticism. But I respectfully disagree that it has to be that way.

My view is that agentic coding LLMs are a tool, and like any tool there are good and bad ways of using them. Left unchecked, they will happily generate a mess. But understood, channeled, and wielded with discipline, they can produce genuinely high-quality work. The craft is in learning how they behave — their strengths, their failure modes, where to trust them and where to verify — and in building the right tools and harnesses around them.

That's exactly what I've been doing. Since early access to GPT-4 back in March 2023, I've spent a large share of my free time researching and experimenting with agentic coding LLMs: testing how they reason, mapping their limits, and refining better tools, workflows, and harnesses around them. Protostar is the result. The codebase currently holds over 100,000 highly engineered lines of C# — not generated slop, but carefully directed and maintained code.

I want to be honest about what this means: I would never have had the time to build a project of this scope so quickly on my own. In a very real sense, the LLM acts as an entire small game studio team at my side — and I direct it. This project simply would not exist without that partnership, and I think that's worth celebrating rather than hiding.

Finally, an honest note for you, the visitor. If you hold a strong opinion against AI-assisted development and that's a hard line for you, then this project may not be the one for you to play or contribute to — and that's completely okay. I'd rather be transparent about how Protostar is made than pretend otherwise. For everyone else curious to see what disciplined, agentic coding can actually produce: welcome.