164. Practical Advice for Building Games as a Service
If I sat down over coffee/tea with a CEO or Senior Investor who wanted to make a GaaS, this is the basic information I would want them to know so they don't lose $100M or more.
I’ve been writing a lot of esoteric papers on the nature of existence, our future, even our identities. This is not one of those papers. Here I try to give the most practical advice possible to game development leaders.
Why GaaS is Popular Today
There is this belief in higher circles of game development that it makes more sense to make a game that you can sell over and over again for the next 10 years than to try to keep making 40 hour games that you can only charge for once. This is a rational belief. This is why I only make GaaS and my GaaS games that have made it to market have had exceptional (12+ year) lifespans. I aim for a minimum of 5 years and with proper maintenance and content additions this can be extended to even 20 years.
Games that can last up to 20 years essentially last for a real world human generation. I will refer to these sorts of designs as Generation Games here.
So with perhaps 90% of AAA development budgets going to GaaS projects, why is AAA imploding? The simple answer is that teams that did not know how to make GaaS were given huge budgets and 6+ years to make games that were dead on arrival (DOA) and had to be shut down almost immediately. Concord will be the classic example. This is the worst case scenario for any developer.
So What Went Wrong?
A lot of things went wrong. The industry moved towards a data driven approach to producing creative content, which should have raised red flags immediately. Unfortunately, the industry (and investors) is often led by business people who don’t sufficiently understand the creative process. I’m a scientist that supra-educated in order to solve “unsolvable” problems in the interactive media industry from 2005 to 2010. I would be forgiven for thinking that I could solve any problem with science and thinking that we don’t need creatives anymore. But I don’t think this way and I well understand that a balanced approach is typically the best approach to all things in life.
Leadership
Before I get into all the technological necessities that AAA have overlooked, I think it makes sense to address the most fundamental ingredient in making great games:
Having a team that works well together. This can happen by accident in small teams where leadership is generally shared. But GaaS can’t be built by a small team, there’s just too much going on. The work can be split across multiple remote teams, but they still need to get along. That isn’t the responsibility of the rank and file, it’s the primary responsibility of Leadership. Organisation is not the primary responsibility of Leadership. That’s what we have Producers for.
Producers often have above average leadership skills and can help fill a Leadership vacuum. But just as often a Producer might be in that role due to technical acumen or because they are trusted by Leadership. This does not magically confer leadership skills upon the Producer. With the possible exception of hyper specialized skilled labor (like myself), Leadership skills are some of the rarest skills in the game development industry. This is just to be expected since game developers are usually gamers and gamers often don’t come from backgrounds that teach strong leadership skills (like competitive sports or the military). Politics does not often generate strong leaders, it should be noted. At least not in modern generations.
If we are having tea (or perhaps coffee for you, I’m straight edge and don’t do any drugs) and you are an investor, then I’m going to tell you that the Leadership asking you for money is the most important criteria for whether you should invest in a project. I’ve had multiple studios with weak leadership hire me to convince investors that they should cough up tens of millions of dollars. And they usually do after talking to me, at least in the web2 space. [Web3 investors often want to make a very quick buck and that’s not compatible with Generation Game development. So I will reject the wrong Web3 investors.]
But many times I’ve been removed from these projects once the studios got the investment money. These projects fail 100% of the time, at serious loss to investors. If you as an investor are depending on a non-founder to provide Leadership to a team, you need to quietly create a backdoor line of communication with that person to make sure they can alert you to problems and protect your investment. It should also be a condition that a key Leadership individual should have the authority on a project to exert that Leadership.
When Serb (Sergey Burkatovskiy) sent me to Minsk to apply my tech to the World of Tanks Blitz project he didn’t give me any details before I arrived. I was entirely loyal and didn’t hesitate when ordered to fly to Belarus in the middle of the Winter at the end of 2013. In the very first meeting he told the team that he was giving me full authority over the project. No one was more surprised than me, but this was a critically smart move on the part of Serb. Without that authority I honestly don’t know how WotB would have turned out. Even though that authority included the power to delay our production timeline, I didn’t delay. I trusted my team and boosted their morale just as I did on the elite sports teams I’ve led. My design work and tech implementation only took 11 weeks. Not 6 years. That’s one of the primary characteristics of the tech I’ve invented.
The projects I’ve been on with ambiguous authority have all crashed and burned. I would go as far as to say that’s why none of my projects have made it to market since World of Warships, which only took me 14 weeks to design. For a project to make it to market, especially a project of this complexity, it’s going to need strong Leadership.
Note that a strong Leader also knows when to kill a project they are assigned to, even if that might hurt the leader making that decision. Leadership isn’t easy, but the Leader needs to protect their Master. A “yes person” is not a strong Leader.
Now if I’m having tea with you and you are the CEO (a more common event for me) then this conversation gets a bit harder. If you are a strong Leader, and that will be obvious if that is or isn’t true if we are having tea, then my job (especially if I’m getting paid) is to protect you and make sure you know the pros and cons of every scenario that might occur within the sphere of my domain expertise.
But if you lack the necessary Leadership skills (which is more common) that’s okay. The CEO should know how to delegate. Ideally they also understand and acknowledge their strengths and weaknesses. Viktor Kislyi founded Wargaming in 1998, but he wasn’t the strongest Leader. He did manage to surround himself with people who were, and give those people sufficient authority to lead their projects. If you are a CEO in a similar situation, you need to pick your lieutenants carefully. This will make or break your tenure. If your Lieutenants are all “yes people”, you could end up making the next Concord.
GaaS is Changing
Not only is technology changing, but so are consumers. If you wanted to make an automobile in 2025 and decided to copy a car from 2013, how successful do you think that would be? As crazy as that sounds, American auto makers did exactly that for decades, relying on protectionism to delay the need for expensive R&D and innovation. The result is that America lost their technological edge and most of their market share.
But in game development, we keep trying to copy successful games from the 2013 to 2015 era. RIOT and Wargaming succeeded then for the reasons I predicted in my Supremacy Goods microeconomic model from 2012. League of Legends was not yet profitable at the time I wrote SG, and it had not been for the previous 3 years. A number of shooters followed immediately after using Battle Pass meta designs suggested in SG.
None of these games should be made in 2025. Even my World of Warships design is showing its age after 12 years, despite still having a marginal technological edge over its competition (due to the Category 4 monetisation model). Roblox has had some commercial success by doing all sorts of unethical things in their meta design. None of these methods are sustainable and I would not recommend trying to replicate it. Many have asked me to do so and I would not replicate or repair their work.
Even if you have no care for ethical arguments and the threat of regulation and consumer lawsuits, the fact of the matter is that Roblox depends on relatively primitive technology (it was launched in 2006). Much more advanced meta technologies exist now.
Before I get into what those technologies are, I need to address how consumers are changing.
Consumer Needs Have Adapted
Developers and even influencers have engaged in all sorts of gymnastics to explain why all these GaaS products are failing. These two reasons are rational:
Consumers are rejecting the injection of ideology into their games. A poor Leader can force their ideology onto their employees, but consumers are not your employees. You can’t force their behavior. They have organised along this anti ideology axis and react very effectively against it to kill new games that attempt it. Branding them as right wing or left wing is missing the point entirely.
Consumers are rejecting unethical monetisation models. While the ESA geared up to fight regulation in the pipe from the EU (The Digital Fairness Act is coming next year), gamers got tired of waiting. They have gone full vigilante and are dealing with the problem themselves by boycotting consumer antagonistic products. This applies especially to Category 1 and 2 meta designs, and are neutral (but this is still enough to kill your game) about Category 3 designs.
Using more advanced models also have the added benefit of passing the EU’s DFA regulations. Given the small margins that older models generate as perfect competition is realized, the DFA will wipe out all products still relying on these models if they use prohibited methodology. Which they do.
Simultaneously the output of recent neuroscience studies on dopamine receptors in humans suggests that dopamine flooding (the prevailing goal of modern GaaS and social media) triggers feedback systems in the human body that could lead to dopamine intolerance. That means that consumers are getting increasingly uncomfortable when they play dopamine optimized games that they used to really enjoy.
Going Social: Adapting to Adapted Consumers
Dopamine intolerance is caused by dopamine pooling in the absence of oxytocin. Oxytocin is released during positive or negative events involving other people that the consumer is emotionally involved with. Using high tech methods these sorts of bonds can be generated in as little as an hour. “Players” (of all genders) use these psychological techniques to drop the guard of others in dating scenarios. This is decades old tech.
The dual receptor (dopamine/oxytocin) interactions in humans are not old tech. They were discovered in 2024 by laboratory scientists and reported by me on my substack in 2025. What this means is that if you want to keep dopamine flooding, you need to introduce oxytocin triggers into your game design or your design won’t work.
Doing this was clearly a good idea anyways 10 years ago because we know that oxytocin builds trust and reduces consumer resistance to spending. Thus all my designs over the last 10 years have advanced oxytocin generating meta systems. They are also consumer friendly so they pass the DFA (which I helped international regulators produce back in 2013).
Oxytocin (Social Class) Scale
PvV: Player vs. Victim. I first introduced this term ~2001 when I was writing for UnknownPlayer. This means non consensual combat between the initiator and the target (victim). This is a very poor performing model, but was encouraged in early Pay to Win games. Generates maximal threat. EVEOnline also uses this social class, which is the primary reason EVE’s adoption has been limited.
PvP: Combat between players who have consented to that combat. Can generate a lot of adrenaline and dopamine, but that’s increasingly problematic.
Leaderboards: These are a good way to generate feelings of competition without generating significant threat. The sterile nature of a LB doesn’t help your game or players much though, unless they have become invested into each other using one of the higher social classes.
Co-op: If strangers are cooperating this doesn’t have much of a social effect because the players don’t really care about each other. Their success may depend on each other, but this is only a weak connection driver unless contact with these other players repeats. My Wargaming designs that you got a chance to see are all in this category.
Strong Co-op: If you play a Friendslop game with your friends, then you are in this category. As a developer that’s ideal if you can sell a game like this to people who have friends, but it’s not particularly good at promoting new friendships.
Massively Multiplayer Co-op: Here players are on large teams (40+ players) where everyone has to cooperate to win/progress. Meta design systems that promote interconnection and interdependence are critical, otherwise you just end up with a Massively Singleplayer game, which could be Class 0 or even lower. My Sid Meier’s Civilization Online (2013), Project Drakara (2011+), StarGarden (2024), and 3 projects at Wargaming (2015) were Class 3 games. None of those products made it to market for production reasons unrelated to the design. I’m not going to lie, that’s been a bit frustrating.
Neurochemical Feedback: I had a team of allied neuroscientists greenlit to go to start secret testing on this technology at Wargaming in 2014. But then the Croatian invasion happened. Sanctions locked my research budget. In 2014 the idea was to measure and optimize oxytocin and dopamine delivery in our products at Wargaming. This probably would come to a shock even to those working at WG since this project was so top secret. Needles were required in 2014.
As of a few years ago my faithful neuro allies developed the tech to the level where it is non invasive. You basically just wear a watch. StarGarden was designed to be a platform to test Class 4 social systems. We were hoping to combine our technologies on the StarGarden project but all the mayhem in web3 made funding difficult. This tech will probably be deployed in a web2 environment first as it is not dependent on blockchain.
The modern versions don’t just benefit the developer. Multiple players can see their nervous systems go in sync with each other in real time. We can even score them on their ability to inter-sync in addition to their analogue gameplay performance. This is entertainment on a whole new level. It could also have applications to dating, or even as lie detectors. That last use I would consider “dual use” and I’m not thrilled about that. Alas, anything with a military application gets funded, so it’s going to happen somewhere.
The world’s largest companies are aware of the existence of this tech, or at least the watch. To date those companies haven’t figured out what to do with this amazing technology. The AI-induced tech tunnel vision could delay its implementation for a while.
Helldivers 2
The best GaaS that’s come out recently was Helldivers 2. The reason for its success, besides being a really high quality creative effort, is that it uses a Category 3 metadesign paired with a Class 1 social system. It is competing against older products that suffer perfect competition running Category 3 and Class -2 or -1. This is a modest increment in design methodology/tech, but when a market is suffering perfect competition, it only takes a small advance in technology to render those previously perfectly competitive products non-competitive.
I can see from their use of “Major Orders” that Arrowhead Game Studios intuitively understood that large scale cooperative meta designs would significantly outperform small scale cooperative games. I would describe their progress as about 10% of the way towards a full Class 3 product. The first 10% that is achieved is generally the easiest and most obvious increment. Each additional 10% progress towards 100% gets increasingly more expensive, complex, and time consuming to implement. Thus without access to the science/tech from a 3rd party the incremental process to a full Class 3 game could take 20+ years.
Thanks for the Tea!
So before we head our separate ways, the real question for you (as either a CEO or investor) would be what do you want to build and why? If you want a game/platform that can generate revenue for 10 or even 20 years, do you have a team that can handle that? Are your leadership systems satisfactory? Building a new team for that would require an amazing amount of work to make sure everyone on the team is sufficiently compatible. Leadership lowers the risk and cost of that process.
Then you have to decide how high tech you want your social systems to be. Social Class is a profit multiplier, and if you build at Class 2 and a competitor deploys Class 3, your window of competitiveness might be cut short. Right now I’d describe any GaaS at Category 4 or less and Social Class 1 or less to be vulnerable to a higher meta tech competitor. Category 5 metas are designed to be at least Class 2, and Category 6 metas are designed to support Class 3 and 4 models.
With some basic research completed even more advanced (and presumably more expensive) metas and classes could be designed. Whether that’s a good idea would depend on how it fits with AI development as that determines available resources and what we want to do with all of those idle workers. Do you want to keep them entertained? How will they pay if they have no income? These are basic questions that should have been answered already. Since AI developers haven’t told us, I assume we aren’t going to be thrilled with their plan, if they even have one.
My view is that if we are going to keep a lot of idle people around, then we need to keep them entertained or they will agitate. Thus, if we don’t plan to get rid of them, and want to keep them very distracted at the lowest possible price point, this sort of scaling tech will be a necessity. Let’s talk about it over tea!