Building Kurtuluş 1919: How AI Helped Turn a 25-Year Dream Into a Strategy Game

Many software developers have a long-term project that stays in the background for years, waiting for the right time and tools. For Coyotiv CEO Armağan Amcalar, that project was Kurtuluş 1919.

For over two decades, Amcalar planned to create a Risk-style strategy game centered around the Turkish War of Independence. The premise was straightforward: players are tasked with defending the 30 regions along the National Pact’s borders against multiple occupying forces.

Over the years, Amcalar made several attempts to bring this game to life. He collected historical maps, collaborated with university history professors to extract accurate provincial data, and wrote early pieces of code. However, the scale of the project delayed its completion.

That all changed recently. Fueled by modern technological advancements, Amcalar finally completed the project over a highly productive three-month period.

The Catalyst: AI and Vibe Coding

The turning point arrived with the evolution of modern AI and vibe coding. While Amcalar designed the game’s core architecture, AI tools were used to handle the bulk of asset generation that often slows down solo indie developers.

AI was utilized across almost every creative asset in the game. Alongside his colleague, Amcalar used AI to generate custom artwork and a complete 13-track original soundtrack. The soundtrack features an opening track titled The First Shot, which adapts the notes of the Turkish national anthem into a different tempo.

Reflecting on the emotional weight of this release during a recent YouTube livestream, Amcalar stated:

“This is the very beginning of an adventure I have dreamed of for 20–25 years.”

Architecture: A Declarative Engine, Not an If-Else Maze

On the surface, Kurtuluş 1919 might appear to be a standard browser-based JavaScript game. However, an incredibly intricate, highly dynamic game engine powers it behind the scenes.

When designing the underlying architecture, Amcalar specifically wanted to avoid a messy, imperative codebase cluttered with rigid “if-else” statements. His inspiration came from an unexpected piece of computing history: Lale Savaşçıları, a classic two-disk 1993 Amiga game developed in Turkey. That retro title featured a dynamic engine driven purely by instructions, allowing the entire narrative and logic to shift based on a few variable changes.

Amcalar applied this exact declarative philosophy to Kurtuluş 1919. He built a comprehensive, JSON-based event registry where every historical event, troop movement limitation, and territorial bonus is defined dynamically.

“Behind this browser game runs an incredibly detailed motor. None of it is hardcoded. If we want to change something, we can alter a simple line of information in the JSON and completely change how rules connect and affect the game.”

Because of this architectural decision, the game is essentially a standalone, versatile engine. By simply swapping out the SVG map and editing the JSON file, the open-source community could use this exact architecture to create entirely different strategy games — whether a historical simulation of the Anatolian Beyliks, or even a political strategy game set in the fictional universe of Game of Thrones.

Weaving Authentic History into Gameplay

One of the most challenging yet rewarding aspects of development was weaving real historical events into the core gameplay loop. The game tracks dozens of historical milestones, including the Erzurum and Sivas Congresses, the Treaty of Sèvres, the opening of the Grand National Assembly, the National Obligation Orders, Soviet assistance, and the Battles of İnönü — all interconnected through a massive, branching web of probabilities and conditions.

For example: successfully opening the Turkish Grand National Assembly requires the player to hold Ankara. If an enemy captures Ankara, the declarative engine dynamically retries for three turns. If Ankara remains occupied after these attempts, the system gracefully adapts by checking if Sivas is under player control, allowing the assembly to open there instead. These events trigger complex chain reactions — if the assembly isn’t active, the historical Çerkez Ethem rebellion simply will not occur because there is no established authority for the forces to rebel against.

Combat mechanics are equally rigorous. Players must defend the homeland against six distinct, AI-controlled enemy nations: Britain, France, Italy, Greece, Armenia, and Bulgaria. These factions don’t simply compete for territory — their behavior reflects different strategic priorities. According to Amcalar, neural-network models for these factions were trained through roughly 800,000 simulated games. Tactical rules — such as a strict maximum of three attacks per turn, and troops gaining “entrenchment” bonuses the longer they remain stationary — force players to think strategically rather than relying on endless blitzkrieg tactics.

Custom UI, Dynamic Typography, and Layered Audio

Building the user interface and audio systems required bespoke engineering solutions. The game map is built on a modern SVG foundation, enhanced with custom shaders and dynamic shadows to create a tactile, 3D watercolor aesthetic.

Typography presented a highly specific visual challenge. Map-based strategy games require regional names to fit perfectly within uneven borders. To solve this, Amcalar developed a custom dynamic typography engine that algorithmically calculates the best position and font size to fit text inside irregular SVG shapes — a tool he noted could be open-sourced independently.

Audio presented another significant hurdle. Standard sound libraries often feel disconnected, so Amcalar built a custom, multi-layer audio generation studio leveraging the ElevenLabs API. Because generative AI models struggle to create multiple distinct, overlapping sounds from a single prompt, this custom interface allowed him to generate individual sound effects and layer them seamlessly into a unified audio file.

AI also played a major role in writing the software itself. Crucially, Amcalar distinguishes that process from simply prompting a model and accepting whatever code it produces:

“I had AI write the code, but I went through every part of it in detail. The design, the decisions, the architecture and how everything should work were mine.”

An Open-Source Legacy

Today, Kurtuluş 1919 is completely open-source and freely available on GitHub, and Amcalar sees the existing version as a starting point rather than a finished destination. The project includes an extensive unit-test network, while its underlying architecture was deliberately designed to make new scenarios and mechanics easier to add.

It stands not just as a deeply engaging historical strategy game, but as a robust software framework — and a testament to what is possible when expert software engineering meets the capabilities of modern AI tools.

AI did not come up with the idea, decide what the game should be, or remove the need for engineering decisions. It lowered enough barriers between an idea and a working product for someone to finally build it.

The whole project has resulted in a complex, declarative engine capable of much more than its current scope. The groundwork has been laid, and Amcalar’s ultimate hope is that the wider developer community will take this engine, fork it, and build their own scenarios.

To learn more about the project and experience this historical strategy firsthand, visit kurtulus1919.com.