Seed is one of those games that has the potential to either dominate the world, or collapse spectacularly under the weight of its own lofty ambitions.
The scale of it is incredible. The aim is to simulate society across an entire planet – Avesta – where each inhabitant, or Seedling, is designed by a player, given tasks, and then left to semi-autonomously find its way in the world. That means tens of thousands of Seedlings have to be persistently maintained on servers, going about their daily business and interacting with other Seedlings, whether the player responsible for them is logged in or not.
Everything in the game is shaped by the players, who will develop the relationships between Seedlings, set up businesses, create jobs, shape the economy, and enact laws that affect whole populations.
If that kind of grand-scale simulation sounds a little like Eve Online, it’s no coincidence. Two of the co-founders of Klang Games, the Berlin-based company behind Seed, previously worked at Eve maker CCP Games (now Fenris Creations) – CTO Oddur Snær Magnússon and CCO Ívar Emilsson.
The third co-founder, CEO Mundi Vondi – the only one from a non-video-games background – was the one who came up with the “everything simulator” concept. “I was a bit in the arts, and making films, and so forth,” he tells GamesIndustry.biz in an interview at Klang’s offices in Berlin. “And then I run into my best friend and co-founder Ívar.
“He was then working on Eve, which of course was very exciting to me, and we started talking. And then I brought to him this document that I had worked on, which was this everything game idea, and he’s just like, ‘Oh, here we go’. But he kind of entertained it, and we started to talk about all the problems: what would it take if you [wanted to] make Eve more accessible, and all of a sudden you have a million players in there?”
The key thing the Klang founders wanted was to have a fully populated game world at all times – as Vondi notes, in any online game, only a small fraction of users are logged on simultaneously, even at peak hours, meaning many MMO towns are deserted for much of the time. “This is something that we were really eager to reverse, and make the content, the building of the house, be a sort of game in itself.”
Those initial conversations happened over ten years ago, but the technical hurdles were considerable, and investors baulked at the idea of running tens of thousands of AI characters in the cloud. “At the start they were just like, ‘That’s insane. That’s not possible’,” recalls Vondi.
But over the course of a few years, Klang managed to overcome the difficulties of scale, and was able to start preproduction in 2018. “We had some breakthroughs, and got some crazy investors to believe in us, and got a little funding, and we proved it out more and more,” says Vondi. “And slowly, slowly this dream became a reality.”
The metaverse hype at the time helped when it came to attracting funding. “That was right when Mark Zuckerberg had taken a big love for the metaverse and was putting in billions and billions and billions,” recalls Vondi. “And there we were, having worked on something along those lines for way longer than anyone else.”
Steep learning curve
The main aim of Seed is to enable an emergent society that evolves in unexpected ways. “We believe in the future that more than 90% of the game is going to be built by the players,” says Vondi, “and it’s going to be way better than whatever any studio could ever build, because they’re not just building it, they’re also a part of it.”
Klang envisages the players forming into three broad groups. Governors will essentially run the societies, enacting laws and planning the economy, much like the heads of corporations in Eve Online. Entrepreneurs, meanwhile, will set up businesses producing goods and services, and providing jobs. Finally, caretakers will make up the majority of the population, with their main goal being simply to look after their Seedling and find work for them. And they do need looking after – neglecting your Seedling’s needs, which include requirements for shelter, hygiene, and so on, will eventually result in them dying.
Vondi says it’s not hard to attract governors and entrepreneurs, people who are happy to sink countless hours into crafting a society. But attracting the more casual players has been trickier.
Starting from scratch in Seed is a formidable prospect. Creating a settlement is essentially a survival game that requires several players to work together to gather resources, build shelter, and slowly work up through a tech tree towards proper houses.
The online multiplayer requirement means there’s no option to fast forward building or gathering, and Seedlings need to sleep for at least a portion of the day – which basically means watching them do nothing for perhaps 15 minutes. In a demo at Klang’s offices, after nearly two hours, our group of three had only managed to build a couple of tents and a crude wooden shower and toilet.
Tamagotchi society
But the game isn’t really meant to be played in this way by the vast majority of users. The idea is that most people will simply plug their Seedlings into fully functioning societies that have already been built by other players. And rather than watching over and directing your Seedling’s every move, the focus is instead on using a calendar to schedule their activities, like going to work or crafting items, only checking in on them from time to time to adjust their schedule as needed.
Essentially, it’s like a massive society of Tamagotchis. This notion is reinforced by the companion mobile app, which works in tandem with the main PC game, and which players can use to check in on their Seedling and adjust various parameters. It also sends notifications that tell you what your Seedling has been up to, or, in the worst case scenario, that they have died.
“The mobile [app] was something we always knew was a critical component of Seed,” says Vondi, who emphasises how important it was to attract more casual players who simply enjoy existing on Avesta.
“It was not that hard for us to make the governance and the business gameplay fun, because there are common patterns we could follow there,” he says. By contrast, the caretakers are the hardest audience to please. “They would always compare this to The Sims or other life simulation games that are single player, highly polished, and tailored around their own individual experience, making sure that there’s a fast forward button, just kind of really cuddling you through.”
Watch on YouTube
There’s a lot of similarities with The Sims in how players remotely manage autonomous Seedlings rather than directly controlling them most of the time. But Vondi acknowledges that it’s a tall order to compete with EA’s rich and many layered world.
“I would often say Seed’s going to be like Eve and the Sims combined, but it won’t have the fidelity of the Sims… But then came our saving grace, which was generative AI. Because all of a sudden, we saw the possibility: wait a minute, we can go far deeper than any simulation game has ever gone, while still maintaining this huge scale.”
“As far as silver bullets go, I mean, that really was the one.”
At any point, the player can send a text message to their Seedling, who will whip out their phone and reply. At one point in our demo, we asked our Seedling to flirt with another Seedling nearby and report back on how it went. You can ask them to do tasks via text, and they will autonomously develop relationships with other Seedlings. It’s somewhat reminiscent of the generative AI experiments that Krafton has been trialling in Inzoi.
It’s easy to see how this might be appealing for more casual players, who might enjoy talking with their Seedling and keeping tabs on their love life. But Vondi says that getting generative AI to work properly in the game was no easy task.
“It was much harder than we thought,” he says. “And because it’s always on, there’s a lot of nuances. If you go offline, your Seedlings still continue talking. How are you supposed to find out what happened when you come back? How are they supposed to react to you being rude to them? How are they supposed to react to one another?”
Dealing with things like interruptions to conversations, and getting the AI to remember things and react accordingly has been particularly tricky. “Like if I tell you it’s my birthday today and then I tell you again tomorrow, you need a very different reaction, [because] one of those days I must have been lying… We started to see a thousand little breakages like that, and then we realised, wait a minute, this isn’t going to be easy at all.”
There’s also the problem of AI hallucinations, where the AI suggests doing things that aren’t actually part of the game. Although Vondi says this has actually helped in a way, since it directed the team towards features that could or should be added. “For example, when we had just started, one Seedling was always saying, ‘We should go for a drink sometime’,” he says – which led the team to actually program functionality where the game would automatically schedule in a time for meeting up for a drink on two Seedlings’ calendars after they had agreed to the rendezvous in a conversation.
The problems and opportunities of a persistent world
The scope of the simulation in Seed is enormous. As Vondi says, if any character walks around a corner in this world, “he doesn’t disappear like in GTA, he’s going to go all the way to his house, and he’s going to brush his teeth before he goes to bed, no matter if somebody’s looking or not.”
Getting to this point has required an enormous amount of technical ingenuity. “When people see Seed, and they start to think ‘How can we do something similar?’, they’ll realise how insane the tech is behind what’s happening there,” says Vondi.
But it’s also expensive. Constantly running a huge simulation in the cloud, as well as including generative AI functionality, does not come cheap. So how is Klang going to make the numbers work?
Vondi says that they’re going to keep it simple to start with. “We’ve tried a few avenues, but what we’re really landing on is just the straight up premium: you buy it, you play it forever.”
He notes that server costs have shrunk dramatically in the time that Seed has been in development. “The race to the bottom of the server centres is real. Prices have come down to a place where you can run a game like this, and that’s really lucky for us, because we bet on that, and that turns out to be true.”
The monetization model also scales. Every player will receive one Seedling, which will be replaced by another for free if it dies, but there’s also the option to buy more Seedling slots, or sign up to a subscription. Vondi adds that there will be cosmetics for sale.
The game still has the option to grind to receive more Seedlings, though. And Vondi says players may even be rewarded with baby pills. “It’s a reverse contraceptive,” he explains. “Everybody’s basically infertile until they take this fertility pill.”
There are some wild ideas in Seed. And it’s difficult to imagine anything with this grand scale and experimental nature being greenlit now, in a post-COVID world where major publishers have hunkered down with safe bets and sequels in the face of spiralling development costs, reluctant investors, and the end of cheap money. It’s arguably a relic from the more optimistic gaming landscape of a decade ago, when the numbers were going up and anything seemed possible.
Seed is launching in early access this summer at https://seed.game, with a release on Steam to follow later in the year. By its very nature, it will require a critical mass of both hardcore and casual players to function. It will be fascinating to see how things play out: whether it will dominate the world, or collapse spectacularly under the weight of its own lofty ambitions.
Klang Games paid for the author’s travel and accommodation for their visit to the Berlin studio.
