Studio Atelico, the game studio building new AI-first gaming experiences, has closed a $5 million seed round.
The company’s mission is to craft AI-first games and technologies that push the boundaries of conventional gameplay, creating new genres that captivate and surprise players in profound ways.
The money came from Air Street Capital and angel investors including Chris Ré (Factory), Thomas Wolf (Hugging Face) and Alex Ratner (Snorkel).
“We’re passionate gamers at heart and we believe that great games are made by people, for people,” said Piero Molino, CEO of Studio Atelico. “This investment accelerates our work in empowering developer creativity and delighting players with experiences that resonate on a profound level. We’re here to create gameplay the world has never seen, and to remind us all of why we fell in love with games in the first place.”
The company formed in December and it raised the funding a couple of months ago. The company has six people. But it still managed to raise a lot of money for its size.
Building games and developer platforms
The studio is committed to protecting and enhancing the rights of artists as they build out their AI technology, Molino said in an interview with GamesBeat. The seed funding will be used to achieve three goals.
It will create games utilizing Atelico’s proprietary technology. It is also is releasing games that exemplify the power of generative AI to unlock innovative gameplay.
Atelico will also accelerate the development of the Atelico Engine. It will build and scale technology and tools to enable other game developers to tap into the power of AI without breaking the bank.
And it will be doubling down on its pro-artist ethos. It will do so by developing partnerships with studios and artists who share in the vision of promoting a fair approach to building and utilizing AI technology.
Nathan Benaich, general partner of Air Street Capital and author of the State of AI Report, led the round.
In a statement, he said, “AI-first technology is set to enable entirely new and exciting types of gameplay. The team at Studio Atelico brings together frontier AI expertise with a proven track record in game design – a rare combination that sets them up to create gaming experiences that players haven’t seen before.”
Alongside this financing, the team is bringing on gaming and tech industry veteran Chris Olson, an entrepreneur in the video games space who also previously led Sega’s growth in the mobile gaming sector. At Studio Atelico, Olson will leverage his broad expertise to drive and accelerate game development and marketing as the studio enters its next phase of growth.
Why AI?
Asked what the motivation was for starting the company, Molino explained he has been working in AI for 15 years now but is a gamer at heart. He has been developing games on the side throughout his career.
“But now I thought it was the right time to start doing it professionally because I sold my previous company that was a tech company building AI infrastructure, and I felt that the technology that I built there, if applied to local devices, would have been the ideal AI technology to build a new wave of games that had mechanics that hadn’t been seen before in gaming,” Molino said. “And so that was the initial inspiration.”
He read a paper from Stanford University researchers looking at generative agents. It talked about a village of 25 characters that were AI governed. They interacted with each other and with the world, creating emergent behavior.
“I thought that it was the future of gaming, and that with my technology, I could make that real and practical, because the approach in the paper would require thousands of dollars to run for a day on Cloud APIs, but with the kind of stuff that I’m doing, it basically runs for free on the player’s device. And so that was the original inspiration to make something practical that could create new game styles that are not possible,” Molino said.
“Depending on the game design, designers can adapt the design to the specific needs of the specific tasks in the game,” said Molino.
With these tools, Molino’s company is creating prototyping tools used for creation. Just a handful of companies are doing this kind of work now.
“We are all discovering what the patterns are,” he said. “We created this internal prototyping tool that makes it possible for us to create the core nugget of the gameplay of a game very, very quickly to test it. Over the course of a couple weeks, we were able to create 10 different prototypes.”
He added, “What we discovered is that there is a window from very strict, handcrafted game design to very free form game design that is governed by the LLM. A good example of freeform could be something like AI Dungeon, for instance, where you just type something and the LLM comes up with what happens next.”
He added, “On the other hand, in the opposite, you have a classical game where you handcraft all the rules. And so we found that we are bouncing back and forth in this spectrum to figure out what’s the right in-between spot where you make rules with the freedom that the LLM affords you, and you create experiences that are new but anchored in game design principles.”
Molino thinks there is great potential in the emotional intelligence in psychological games where diplomacy or gamesmanship is happening. An AI game director might detect fear in a player and then adjust the gameplay to address that specific fear. The goal is to get sophisticated AI gameplay to run on a laptop with local AI processing at zero cost.
Ethically sourced games
By ethically sourced, Molino said the company uses small models that can run on devices that don’t have too many processing or memory resources.
“The important thing here is the data that the models are trained on is data that is either publicly available or consensual data,” he said. “Furthermore, on top of these base models that can do many different tasks, similar to what Open AI models do, we adapt them to the specific game.”
This means you can query them to do many different things, but in order to make them work really well, despite the fact that they are small, the team adapts them to the specific game. And so the data for art, graphics, narrative, characters are all sourced from the creators of the game themselves, and you need substantially less data to do that.
He said, “You have ethically sourced models with your own data that adapts the model to the specific game and use case. And the combination of these two ends up meaning that everybody that was involved with the process of creating these models is rewarded,” he said.
Making efficient AI models that run on the device
The problem of sending all AI processing to the cloud is that there’s a logjam that can occur and slow down the processing. And it gets a lot more expensive.
To build a model, Molino said you have to start with a model trained on a ton of data — trillions of tokens, collected in data sets thare are openly available and ethically sourced. But small models that run on a device are not as good as the large models that are available through the cloud.
That’s where the company takes a second step, which is the adaptation to the domain, which bridges the gap in quality. In this case, the owners of the data are developers who are creating the games. The effect is reducing the amount of storage required for the LLM. The team adds adapters with small pieces of models. These adapters are loaded depending on the task. If you need to generate the plan of the day for a character, then there’s a small adapter capable of doing just that. You load that, get the plan, and then create the dialogue lines for that specific character. You load an adapter for that character. These small models are maybe tens of megabytes. They load on the fly, and make the solution work out without consumer a lot of AI processing power.
This is one way that AI processing has gotten less expensive over the last three years. But depending on the specific task, you need more or less intelligence, and so you need smaller or larger models. As much as possible, it runs on one device.
Making the impossible playable
Studio Atelico’s mission is to build games where AI makes the impossible playable. AI in games will be used in service of creativity, helping empower artists to explore new heights. Studio Atelico wants to create worlds that change and react in real-time, offering new horizons in gameplay while keeping the heart and soul of the creator’s vision at the forefront.
Just like physics engines inspired flight simulators and driving games, Molino expects there to be an explosion of AI games that result from the new AI tools. The goal is to make games with AI that could not otherwise be made.
“We are not interested in making development 10%, 20%, 30% faster or cheaper. We are not specifically interested in taking away jobs or anything like that. Our interest is in providing new tools that create new experiences,” Molino said.
The company plans to create NPCs with lifelike intelligence that remember your actions and learn autonomously over time, like in GARP, its tech demo implementing a realtime version of the generative agents paper. It will also do real‑time procedural generation of environments and even rule sets, adapting dynamically to each player.
It is also creating new interaction systems-negotiation and persuasion mechanics, inspiring or rallying allies, and on-the-fly improvisation. And it is creating emergent quests, where agents generate missions based on the evolving world state and your prior choices.
And Molino said it is driving emotion‑based group behaviors, with NPC moods spreading and affecting gameplay flow.
Studio Atelico said it celebrates every player’s imagination and fuels every developer’s vision, all without breaking the bank on Cloud AI. The company’s first game will be released soon.