Supercell London is leading development of a new title not aimed at the “typical Supercell demographic”, says former Space Ape and Beatstar lead Charmie Kim.
Four months after Space Ape officially became Supercell London, Kim is now leading a small team in the firm’s Spark program. Her team has also dabbled in using AI NPCs – but it should not be considered an ‘AI game’, says Kim.
It’s all part of what might be termed Supercell’s new ‘experimental era’, after the unconventional launches of Squad Busters and Mo Co.
We were able to ask Kim a few questions over email ahead of the opening mobile session at Develop today. The below Q&A has been edited for readability and clarity.
I know it’s under wraps, but what can you tell us about what you’re working on right now…?
We are in early prototyping stages so anything I say would be wildly inaccurate in a few months time anyway. I can say that we are part of the Spark program and arguably the furthest along, and we are targeting a demo that is not your typical Supercell demographic.
I know this is going to sound vacuous without more detail but innovation is at the heart of what we’re trying to do. My personal philosophy towards innovation is one that is radically player-centric, I don’t believe in innovating for innovation’s sake. So we’ve done a lot of concept testing and market analysis before even deciding on the game to make.
Supercell’s CEO recently called for more risk-taking – how has that affected your thinking?
Supercell has always placed huge importance on taking risks and doing something new. When I was leading Beatstar and we were trying to reimagine music games on mobile, those values were what we were following as well. Supercell games were always a north star for me my entire career. So in that sense I’m not sure anything’s changed.
But where I might add to what Ilkka said is within the mobile I think we could do better celebrating innovations, and that might spur on more innovation.
I find a lot of innovation in games like Homescapes and Whiteout Survival that sometimes my fellow developers will look down on. I think it’s ironically because people tend to equate familiarity with quality and think of that as innovation.
If the ‘innovation’ doesn’t look like the games they’re used to playing on console or PC, likely as teenagers, then they have a hard time recognising it as innovation. That’s been my experience, anyway.
What has come out of your AI Labs and Spark program? Has that helped you think differently about building new games?
Very timely question because our team worked very closely with an AI Lab team Starward for an AI NPC experiment very recently. Two of our NPCs had their dialogue being generated in real time based on the player’s game state and the NPC’s personality traits.
I wish I could go into more detail but it was very much just an experiment and I don’t want you to write a headline like ‘Supercell’s new game is an AI game…!’.
In fact the conclusion for us was because we’re not an AI game at our core and we need to sort out our core loop first, it was not going to be our focus for the time being. That said I’m fairly certain we will come back to it once our core loop is more fleshed out. And when we get there we will know a lot better now how we want to approach it.
We found out, for example, that the realtime aspect of the LLM generation was not that desirable for us, or even too risky. The right approach would be to send the LLM a dump of the player’s game state on a daily basis, generate the dialogue, and then we can run other AI to do things like checks for bad words and check for quality before it goes back to the player.
Otto who does a fantastic job running AI Lab is chatting quite a lot now to Stephan who runs Spark to see what other or further collaborations we could be having there.
Famously, Supercell doesn’t release many games, but that seems to have changed lately with Squad Busters and Mo Co. Is the company in a more experimental era now? Is there no longer that pressure to deliver a mega-hit every time?
The pressure is always there but the approach to how games get there has certainly been more experimental. I think Squad took the approach of a big brand marketing launch because the game was using multiple Supercell IPs.
Mo Co is doing a completely different approach of slowly growing the game in ‘launch’. I think we can expect future games to continue to try different things.
Again – famously – in the past Supercell’s dev teams decide whether or not to go global or kill a project. Does that process continue today at Supercell London, and do you know if that has changed over the years as teams and expectations have grown?
It definitely is still the case. It continues to be held as a fundamental value at Supercell that teams ultimately have full autonomy. It’s so integral to the studio and so dear to Ilkka’s heart, I can’t see it ever changing.
What do you think the Space Ape teams have brought to the Supercell culture in the transition from Space Ape to Supercell London?
It’s only been four months since the opening of Supercell London so it’s still fresh, we’re still in the thick of any changes, and I think it’s not linearly felt.
Space Ape always had a strong culture of sharing and communication, and a lot of that I think we’re seeing carried into Supercell London in the form of certain rituals and hangouts. But also we’re being enriched by the new people joining Supercell London without any Space Ape hang-ups.
Supercell London is now Supercell’s second biggest studio with 85 people, and a big portion of us came from Space Ape. Helsinki is about six times our size. 16% of culture is going to be significant no matter what, but because people are spread out across different games and central tech teams the impact on the whole org is hard to see.
Speaking for our team, I think we are bringing some of Space Ape’s new game practices into Spark in a positive way.
Okay let’s do the AI question…how have you used these tools in your processes? Are there any specific tools you’d recommend to other devs working up new games?
Apart from the AI Lab experiment we’ve done, the whole team and pretty much all the disciplines are now using AI in various ways. During our concepting phase we used a lot of Midjourney for exploring different themes. I had ChatGPT analysing a lot of market data for me to gather insight.
Within the last month or so I’ve seen a gigantic shift in the way our programmers are talking about AI for code. Cursor AI seems to have fundamentally shifted the way they’re programming.
One of the central tech teams have built us an AI bot to answer our questions on Supercell code whilst they’re trying to be on holiday, because we’re in the middle of transitioning our Space Ape tech stack over to theirs. Making quick sense of legacy code I think is potentially a game changer for teams that want to move fast within big companies.
