Inside Prologue: Go Wayback, PUBG Creator’s Bold Survival Experiment

Prologue: Go Wayback is the kind of project that doesn’t arrive quietly. It’s the debut game from PlayerUnknown Productions, the studio founded by PUBG creator Brendan “PlayerUnknown” Greene, and it’s positioned as the first tangible step toward an enormous long-term vision known as Project Artemis. That vision: a vast, persistent, simulation-driven world that Greene has famously compared to a kind of “realistic Minecraft” for potentially millions of players.

Rather than trying to build that colossal world in one leap, Prologue: Go Wayback focuses on a much tighter experience: a single-player, roguelike survival game with a strong emphasis on harsh weather, emergent exploration, and—most intriguingly—machine learning–driven terrain generation that happens locally on each player’s PC. Every time you start a run, the game produces a new 64km² wilderness for you to cross, and no two journeys are meant to be the same.

From Battle Royale Pioneer to Immersive Survival Worlds

Brendan Greene is best known for leading the design of PlayerUnknown’s Battlegrounds, the breakout battle royale that popularized the last-player-standing formula and helped cement the genre across PC and consoles. PUBG’s impact on multiplayer shooters is well documented, with the game reaching tens of millions of players and helping push the broader battle royale trend into the mainstream, alongside titles like Fortnite and Apex Legends. Major outlets such as the BBC have noted PUBG’s key role in the battle royale boom and its massive global player base.¹

After PUBG’s success, Greene shifted his focus away from competitive shooters toward something more experimental: large-scale, simulated open worlds with deep systemic behavior, dynamic environments, and near-infinite replayability. PlayerUnknown Productions was established as a separate studio to pursue that goal, distinct from current PUBG developer Krafton.

Prologue: Go Wayback is explicitly described as one of three stepping-stone projects leading up to Project Artemis. Rather than a multiplayer sandbox, it’s a focused survival experience that allows the team to:

• Prototype its machine learning terrain generation pipeline.
• Test systemic survival mechanics at scale.
• Study how players move through and interpret naturally shaped landscapes.
• Iterate quickly in a smaller, single-player context before committing to huge infrastructure for a multiplayer “super-game.”

Core Concept: A Roguelike Survival Trek Across a Vast Wilderness

At its heart, Prologue: Go Wayback is about a straightforward but demanding objective: surviving a journey across a giant, dynamically generated wilderness from a starting point to a distant weather tower. Each run drops you into a remote cabin somewhere in this world; from there, every decision—what path you take, which resources you gather, when you rest, and how you handle the weather—can determine whether you make it to the tower or perish on the way.

Key characteristics of the game loop include:

• Single-player focus: No other human players, just you against the environment.
• Roguelike structure: Each run starts fresh on a new map; death is a hard reset, but knowledge and skill carry over.
• Open-world exploration: There’s no critical path; you choose your route and pace.
• Environmental storytelling: Points of interest, ruins, and natural formations create light narrative context without a heavy scripted story.

By keeping the goals simple and the world lethal, the team puts emphasis on emergent gameplay—the unscripted stories that arise from the interplay between systems, environment, and player choice. This design philosophy aligns closely with contemporary views in game design research that stress systemic interaction as a driver of replay value and player engagement.²

How Machine Learning Shapes Every New World

The most unusual aspect of Prologue: Go Wayback is the way it generates its terrain. Instead of relying solely on traditional noise-based procedural generation, PlayerUnknown Productions is building an in-house machine learning (ML) model that works as a tool for their artists and level designers. The model is trained on publicly available, open-source geographic data that the studio vets to avoid using copyrighted material, then used to produce realistic base terrain maps inside Unreal Engine 5.

In practical terms, this process looks roughly like:

1. The ML model ingests curated heightmaps and landscape data, learning patterns of natural terrain such as valleys, ridges, river basins, and mountain ranges.
2. When a new run starts, the game instructs the model to generate a massive 64km² base terrain layout on the player’s PC.
3. Once the heightmap exists, Unreal Engine 5’s procedural tools populate it with assets—trees, rocks, lakes, rivers, and other environmental features—guided by rules authored by the art and design teams.
4. Final passes add gameplay-relevant points of interest and resources, layering interactivity on top of the naturalistic base.

The result is intended to feel less like a random noise function and more like hiking through a plausible, cohesive landscape. Researchers in procedural content generation have highlighted that blending machine learning with designer constraints can produce richer and more believable spaces than purely random or purely hand-crafted approaches, particularly at large scales.³

Crucially, the studio emphasizes that map generation happens offline on the player’s own machine, not via cloud-based AI services. This has design and privacy implications: it removes dependencies on external servers and keeps the heavy lifting local, more in line with a classic PC game runtime than a live-service platform.

Survival Mechanics: Weather, Warmth, and Resource Scarcity

Prologue: Go Wayback is designed as a hardcore survival experience rather than a laid-back hiking simulator. The trek across the map is dangerous primarily because of environmental hazards and resource scarcity. Core survival systems include:

• Body temperature: Staying warm is central to survival, especially at night or in storms. You’ll constantly need to seek shelter, craft fires, or layer clothing.
• Hunger and thirst: Food and water are limited, forcing trade-offs between detouring for supplies and making direct progress toward your goal.
• Fatigue and pacing: Long marches may require rest, and mismanaging stamina can leave you exposed to the elements at the worst possible time.
• Dynamic weather: Weather events can force last-minute route changes or turn safe areas into lethal zones.

This emphasis on environmental survival echoes broader trends in the genre, seen in games like Rust, The Long Dark, and DayZ, where cold, hunger, and exposure are often deadlier than enemies. It also resonates with academic work in survival game design, which underscores that meaningful tension often comes from managing overlapping resource pressures rather than from combat alone.²

Roguelike DNA: Failure as Progress

Labeling Prologue: Go Wayback a “roguelike” isn’t just marketing shorthand. The game inherits several key traits from classic and modern roguelikes:

• Procedural generation: New world each run, reshaping routes, resources, and terrain.
• Permadeath: When you fail, your character and run-specific progress are gone.
• Emergent challenge: Difficulty comes from systemic interactions rather than scripted difficulty spikes.
• Learning curve: Player skill and knowledge—about biomes, weather patterns, and navigation—grow over time, even as the environment resets.

Designers and researchers have noted that this “fail forward” model can strongly encourage replay and experimentation: players are more willing to take risks and learn because death is expected and framed as part of the experience, not a punishment for having played “wrong.”³

Tools for Players: Map Editing and Custom Worlds

Beyond the automatically generated worlds, Prologue: Go Wayback includes a map editor that lets players build or customize their own landscapes. One particularly striking feature is the ability to import an image as the basis of a map. In practice, this means players can:

• Sketch rough layouts and let the ML pipeline embellish them into full terrains.
• Use stylized or real-world heightmaps as seeds for new wildernesses.
• Share creative layouts with others, depending on how the editor tools evolve over Early Access.

This aligns with a broader industry trend in which studios increasingly provide players with creation tools, blurring the line between developer and audience. Titles like Dreams and Roblox have shown how player-authored content can extend a game’s lifespan dramatically, while many PC titles rely heavily on modding communities. Valve, for instance, has long acknowledged how modding in games like Half-Life and Skyrim has fueled extended engagement and even birthed entirely new genres.⁴

Ethical and Legal Considerations of Machine Learning Content

PlayerUnknown Productions has publicly emphasized that its ML model uses vetted open-source data and is designed to avoid pulling from copyrighted datasets. This mirrors the growing scrutiny over how game studios and other media creators train their AI systems. In 2024, the European Union adopted new AI regulations that place special emphasis on transparency, training data provenance, and accountability for AI models used in commercial products.⁵

By clarifying that their terrain model:

• Relies on publicly available geographic data, not scraped copyrighted artwork;
• Runs locally and offline on the user’s PC;
• Is used as a tool for artists rather than an autonomous content factory,

the studio positions Prologue: Go Wayback as an example of human-guided AI rather than a purely automated generative system. This approach is likely to be increasingly important as more studios experiment with machine learning while navigating evolving regulation and community expectations.

Early Access: A Long-Term Experiment, Not a Finished Product

Prologue: Go Wayback is launching in Early Access on PC platforms like Steam, with the studio signaling that it expects to remain there for an extended period—potentially around two years, based on reported estimates. During this phase, the team aims to:

• Iterate on survival mechanics and difficulty tuning.
• Gather data on player movement and behavior in large ML-generated maps.
• Refine the terrain generation model and asset population systems.
• Expand features like the weather system, map editor, and world interactions.

Early Access has become a standard model for complex systemic games, particularly in the survival and sandbox genres. Valve’s own analysis of Steam data has found that Early Access can help teams de-risk ambitious projects by allowing them to iterate with real player feedback, provided they communicate clearly and maintain regular updates.⁴

Project Artemis: Toward a Massive, Persistent Survival Sandbox

While Prologue: Go Wayback stands on its own as a survival game, much of the excitement—and scrutiny—comes from its role as a testbed for Project Artemis. Public comments from Greene and the studio frame Artemis as a long-term endeavor to build a huge, contiguous world where millions of players could coexist in a shared environment driven by simulation, not scripts.

Prologue’s contributions to that ambition include:

• Scalable world generation: Demonstrating that realistic terrain can be generated quickly, at large scale, on consumer hardware.
• Player behavior insight: Learning how players navigate, settle, and survive in naturally shaped spaces.
• Systemic survival design: Refining the rules for weather, resources, and environmental risk in preparation for multiplayer contexts.
• Toolchain maturation: Hardening the ML model, procedural placement tools, and editor workflows for a much larger project.

Whether Project Artemis ultimately reaches its most ambitious goals remains an open question; massively multiplayer, simulation-heavy worlds are notoriously difficult to scale and maintain. Nonetheless, Prologue: Go Wayback is notable as a concrete, playable milestone on the path toward that vision, giving players and observers something tangible to evaluate rather than just concept art and design pitches.

Frequently Asked Questions

Is Prologue: Go Wayback a multiplayer game?

No. The game is designed as a single-player survival roguelike. You won’t encounter other human players in your world; the core challenge comes from the environment and your own decision-making.

How big is each map in Prologue: Go Wayback?

Each run generates a new 64km² (roughly 8×8 km) wilderness. The scale is large enough that choosing efficient routes, reading the terrain, and planning for weather conditions become central to survival.

What makes the machine learning terrain different from typical procedural maps?

The terrain is generated by an ML model trained on real-world geographic data, then refined by designers. This is meant to produce more natural-looking valleys, ridgelines, and watershed structures than purely noise-based procedural generation, while still allowing for infinite variation.

Can players create or customize their own maps?

Yes. A map editor lets players build custom layouts or import images to use as the foundation for new worlds, which are then processed by the game’s terrain pipeline. This gives creative players a way to shape their own challenges.

How does this game connect to Project Artemis?

Prologue: Go Wayback is positioned as the first in a series of projects leading toward Project Artemis. It’s a test environment for machine learning terrain, survival systems, and player behavior analysis, all of which will inform the design of a larger-scale, multiplayer world.

Why Prologue: Go Wayback Matters

In an industry crowded with familiar formulas, Prologue: Go Wayback stands out as a high-risk, high-concept experiment. It combines several frontier ideas—massive ML-generated terrains, offline procedural worlds, and a long-term roadmap toward a persistent survival sandbox—while still anchoring itself in a concrete, understandable premise: survive the trek or die trying.

Whether it ultimately succeeds as a compelling survival game or as a stepping stone to Project Artemis, the project is significant for at least three reasons:

• It showcases a pragmatic use of machine learning in game development: not replacing artists, but augmenting them to build large, varied worlds.
• It treats Early Access as a laboratory for systemic gameplay, inviting players to co-shape the evolution of its survival mechanics and world generation.
• It offers a rare glimpse into the long-term thinking of a major game creator who is willing to step away from a proven formula (battle royale) to pursue something less certain but potentially transformative.

For players interested in survival games, systemic design, or the future of AI-assisted worldbuilding, Prologue: Go Wayback is more than just another title on the release calendar—it’s a live experiment in what the next generation of virtual worlds might look like.