Democratizing Virtual Worlds

Building even the most complex forms of games is becoming capital-efficient

This article is about the astonishing rate of change we’re all observing across the industry: not only the “hot” aspects like generative AI, but all of the forms of automation that are reshaping software development in general—and game development in particular.

All of these are manifestations of the same megatrend of democratization.

As technologies are democratized, they result in an exponential improvement in the capital-efficiency of pursuing an idea. The impact is that far more people within the market get the opportunity to pursue creative ideas.

Here are a few things I’ll cover:

• Tortoise and Hare development strategies—and why the Hare will usually win.

• Why understanding exponential productivity curves is critical to competitiveness

• The opportunity cost of not being ahead of productivity-improvement curves

• How game development may converge to a “creative singularity”

Exponential Productivity Growth

People who are watching changes in artificial intelligence frequently comment how hard it is to keep up with things that seem to be changing daily.

Part of what makes the current rate-of-change so bewildering is that it isn’t simply that the rate of change is high—it is that the rate of change itself is increasing (i.e., the second derivative is increasing).

Let’s consider what this means for productivity changes.

Here is a comparison between what a high rate of change (25%) looks like, starting from a value of 1, compounding over 16 time periods — compared to the same starting point, but in which the underlying rate of change is increasing by 1% per period. Imagine these values being assigned to an individual’s productivity, and you can imagine why people are now talking about the “100X engineer” or the “100X artist” in an organization:

Opportunity Cost

Another way of thinking about the above chart is in terms of opportunity cost.

Consider two competitors. One (the blue area) adopted a set of practices in a market regularly growing by 25%. The other (in orange) used a set of practices where the internalized productivity gains are regularly compounding.

For many of the periods in the beginning, things seem to be going well for both companies. But by the end of the graph, the orange competitor’s advantages seem overwhelming.

For the blue competitor, the orange area represents the opportunity cost of sticking to a set of practices that may already be outdated.

A Year of Innovations in One Week

In 2023, the Game Developers Conference was in full swing again: developers from around the world converged on San Francisco to network, learn and imagine the future of the industry.

The buzz around GDC was mostly related to Epic’s keynote: the use of Unreal Editor as the development platform for Fortnite creators—clearly directed at Roblox; and the retooling of their asset marketplace, Fab, to encompass to all 3D engines, including their main rival—Unity—while taking only a 12% revenue share from developers.

All of this was against a backdrop of a flurry of artificial intelligence news:

• GPT-4 showing “sparks of artificial general intelligence” according to Microsoft;

• Google’s Bard getting incorporated into Google Docs

• Adobe’s Firefly, incorporating generative AI into its line of creative products

• The parallel GTC conference from NVIDIA, where Jensen Huang announced exponential improvements such as a 40X improvement in the time to perform computational lithography (from 2 weeks to 8 hours, an important phase of the chip design/manufacturing process), the availability of NVIDIA-trained foundation models for LLM, images and bioscience.

Marketplaces like Fab, democratized technologies like 3D engines and Fortnite Creator mode—as well as generative AI—are all manifestations of democratization:

• AI is democratizing by automating tasks, making easier for developers to create intelligent and adaptive systems—as well as generating content.

• Automation—whether or not powered by AI—is about democratizing by performing tasks without human intervention. This includes code generation, CI/CD processes, serverless, autoscaling, etc.—or simply the automation of publishing enabled by the likes of Roblox and Fortnite.

• Crowdsourcing democratizes content and code by letting the market build more of the parts you need in a capital efficient manner: open-source collaboration, competitions to develop solutions, and user-generated content.

• Marketplaces democratizes software development by facilitating plug-ins to existing frameworks. This enables reuse of code and content—while promoting higher-quality software by providing access to reliable components.

A day after I returned, I went to Penny Arcade Expo in Boston (PAX East), where you get a different view of the industry: the consumer. And what I saw there had a strong theme of independent development. The largest exhibit was indie game publisher Devolver (several of the large publishers were notably absent).

I think the future looks bright for small, agile teams—whether that’s indie game developers, or capital-efficient teams inside large publishers.

Tale of Two Startups

In today’s startup environment, I’d much rather be a hare than a tortoise.

The hare is agile and fast. The hare automates everything that can be automated, and eliminates time-consuming steps in processes that make developers wait.

The hare achieves product-market-fit faster, and then compounds their early market lead into an “unfair advantage” that redefines what product-market-fit means for its competitors.

A lot of tortoises are going to get caught in the following circumstance.

Despite an earlier start, the tortoise ships products much later than the hare (in contrast to the classic fable).

When the tortoise finally gets to market, their product is relegated to niche-status. Even worse: their unit economics will be inferior, because they’re saddled with compounding technical debt and inefficiencies from building outside their core competency.

Are you a Hare?

The hare automates everything that can be automated:

• Uses of off-the-shelf software in all the cases where solutions already exist: accepts solutions that deliver 80% of requirements today rather than 100% after building for years (in all likelihood, as these off-the-shelf options continue to compound their own R&D, they will deliver more than 100% in the future, faster and more economically than you will)

• Uses generative AI where already possible—and rapidly incorporates genAI for other aspects as it matures.

• Crowdsources as much as possible: e.g., use of open-source software; third-party software organized as platforms and marketplaces to help you leverage a network of developers for your own problems (Unity Asset Store, Epic’s Fab, Beamable Live Services marketplace, etc.) — and when possible, crowdsource from your own customers as well.

• Identifies and eliminates steps in processes that slow down agility; most of these are because there are too many people in a process. For example, if you need to “talk to the server guy” every time you need to iterate your software—then your might be going from minutes or hours per iteration to days or weeks.

• Keeps team size to the minimum required. Challenges themselves to find ways that a 3-10 person team can do what recently required the work of a 30-100 person team.

Scale x Efficiency

Earlier I showed the math for what happens when productivity gains are compounded continuously.

Here’s a graph of two of the most important factors that go into productivity:

Efficiency: the relative cost of an individual task (10 here means 10x as efficient, meaning you could do 1 task with the same level of effort).

Example: if you can create some backend logic for your game by yourself, without requiring a bunch of people to come together and have a conversation, maybe an individual developer can do it at 10% of the cost.

Scale: the ability to simultaneously execute concurrent

Example: maybe an art director in the future will create the style and archetypes for the visuals in a game, and then train their own model based on their own designs—enabling them to scale out production without any real ceiling.

These factors multiply by each other—so you could have 100x the productivity in a situation where each task is 10x as efficient and you can scale out the effort simultaneously through 10x the automated pipelines.

MMOs and Productivity

Massively multiplayer online games (MMOs) are amongst the most complex and complicated types of games that exist. They often combine immersive 3D environments, persistent online worlds and complex economies. These games are the foundation for “the metaverse.”

Let’s take a look at how the combination of off-the-shelf software and automation have transformed the process of developing this market over time.

The Build-Everything Era

Around the time that World of Warcraft was created, you had to create just about everything: your own 3D engine, all of the server technology, all of the middleware. That’s why many tried and failed to compete with WoW—underestimating not only the task of creating the game—but the compounding effect of all the systems, content and audience that WoW was aggregating.

In the following charts, green represents the conceptual/design jobs; the dark blue represent engineering jobs; and the light blue involve content creation jobs.

3D Engine Democratization

The first area of technology that reshaped game development was the emergence of 3D engines that were relatively easy to use: Unreal Engine, Unity, etc. This meant that big chunks of the software no longer needed to be made from scratch.

It’s important to note that while many jobs still remained after 3D engines came onto the scene—this was a highly technical job that was larger than most (i.e., the size of these hexes are not equal). Indeed, at the same time as 3D became easier from an engineering perspective, it caused companies to invest in even greater quantities of 3D content.

Backend Democratization

Over time, cloud-native technologies emerged: containers, orchestration and live services platforms. Similar to how the 3D engine democratized the engineering of the frontend of a game, these technologies are democratizing the backend.

Content Democratization: Creative Singularity?

In the coming years, we may see generative AI and automated content pipelines democratize the process of content creation—again, dramatically reducing the number of jobs within many games:

Could one create an MMO with a handful of people? Perhaps a great systems designer with a vision for how the game will work, an artist who sets the creative direction and trains a fleet of AIs to replicate themself—and then a technologist who knows how to reliably and rigorously glue the pieces together? I think so.

This is the creative singularity: as the cost and number of engineering and content tasks reaches their limit of however-closer-to-zero they can be, the creative idea may someday prevail over the inertia of execution.

But the most important takeaway here is this: if you’re building a game or virtual world—make sure you’re building a game using the democratizing technologies available today… not with the inefficient methods of the past.

Optionality

How does one thrive in a world that is moving at this rate of change?

One thing you can’t know at the outset of a project is how mature technologies will be by the time you start needing them. For example, it would be hard to start a game development project today with the assumption that your entire 3D pipeline will be automated with generative AI. Although these technologies look promising, they aren’t really there yet (although that could change soon after I write this).

The way to protect against being into the “blue” side of the productivity curve I discussed earlier is to make sure you’re incorporating flexibility and adaptability into all of your systems:

• Treat nothing as sacred—be prepared to automate everything that could be automated.

• Avoid building technology dependencies into internal development projects that are likely to be obsolete by the time you ship.

• Build a culture of focus: identify your core competency and stick to it—everything else exists only to support your creative vision.

• Hares need to move together in droves: leverage technology partners that have a similar worldview: they should be focused on their competency; leveraging marketplaces and crowdsourcing to populate the areas they aren’t great at; and focused on speed.

The future is bright for those willing to see the increasing rate-of-change as a competitive advantage to be leveraged—rather than a threat to be feared.

Further Reading

• Cloud Native Worlds is about building games and virtual worlds with the modern serverless, container and interoperability frameworks that avoid reinventing years of market-based solutions for scaling the cloud.

• The Direct From Imagination Era Has Begun is about all the technologies converging together to make “the holodeck.”

• The Four Horsemen of the Videogame Apocalypse is about the forces that are reshaping the game industry.

• The Five Levels of Generative AI for Games is about how one might think about progress and possibilities for GenAI in game development.