"If you're competing on cheap labour then you will get wiped out by AI"
Simon Davis of GOAT Gaming on why AI will overthrow old industry models and what's coming next for Telegram games.
Welcome to AI Gamechangers, your weekly look at how artificial intelligence is reshaping the games industry. We talk to innovators building with generative AI, explore how it's genuinely being used in production, marketing, monetisation, and more, and highlight emerging tools that are worth watching.
In this edition, we chat with Simon Davis, co-founder and CEO of Mighty Bear Games, the studio behind GOAT Gaming. He explains why AI isn’t just a trend – it’s a tidal wave transforming how games are made, who makes them, and how fast they scale. From vibe coding in the office to powerful AI agents that battle for prizes on Telegram, this one’s packed with bold predictions and practical insights.
As always, scroll to the bottom of the page for your round-up of the latest news and links.
Simon Davis, GOAT Gaming
Join us as we get to know Simon Davis, co-founder and CEO of Mighty Bear Games. Based in Singapore, the studio is behind GOAT Gaming, an AI-powered gaming ecosystem on Telegram.
After years in mobile and web3 development, Simon’s team embraced AI to solve a content bottleneck and quickly transformed the entire company.
In this interview, Simon outlines why legacy studios are vulnerable to AI disruption, how his team embraced AI-generated content, and what they’re building next: agents that compete on your behalf, roast you in chat, and could soon transact autonomously using crypto.
Top takeaways from this conversation:
GOAT Gaming climbed to majority AI-generated content in under a year, proving that small teams can scale by integrating AI into the creative pipeline.
Simon encourages vibe coding across the whole company, with even non-technical staff expected to use AI tools to automate workflows.
Simon sees AI as an “extinction-level event” for traditional outsourcing studios, especially those built around low-cost labour, as the marginal cost of content creation collapses.
The team is betting on the synergy between AI and web3, developing autonomous agents that compete in games and transact via crypto rails, creating new monetisation models and gameplay formats.
AI Gamechangers: Let’s break the ice with a little bit of backstory. What’s the elevator pitch for your business?
Simon Davis: GOAT Gaming is a platform on Telegram where players can compete and win real-world prizes.
There have been about five and a half million players to date. We have a number of games. We're adding new games all the time.
Within these games, players have various different competitions and raffles, and they can win either crypto or real-world goods.
In addition to that, we are launching a series of AI agents that players can use to compete on their behalf, and also represent them on socials. We actually unveiled the proof of concept last week: her name is Amy. Amy went live, and she was roasting players in the chat! It was pretty entertaining.
It’s coming together as a very big vision that we have to create the destination for financial gaming, and then adding AI gamification and engagement on top.
How has your company evolved with time? You started out in mobile games, then embraced web3. What sparked this latest pivot?
It actually came out of necessity, and it came a bit earlier than other studios.
In late 2021, we started making a game called Mighty Action Heroes, which is still live. It's our first web3 game, and one of the requirements I set for the team is that every hero should be unique. My co-founder said, “Well, that's 25 million heroes!” And I shrugged and said, “Figure it out.” Because that's what the CEO does, right? You just make unreasonable requests!
He went away and started getting into what was then called “procedurally generated content”, which is basically early generative AI tools and pipelines. We were one of the first teams to start using Stable Diffusion, I think. We started experimenting with these tools pretty early on.
I set the requirement that by the end of 2023, half of our output should be AI-generated. Actually, we exceeded that in May of that year. In five months, we went from a low single-digit percentage to more than half of our content being AI-generated. Today I'd say it's over 95%. Humans are there to guide it and act as directors, certainly on the content side (be it audio or visuals, or even things like gameplay balancing).
“The bigger questions for me are the ethical ones. If you have agents interacting with humans, how do you make sure that's done in a way that's safe? If people choose to interact with agents rather than have friends, what effect does that have on society? What's our responsibility?”
Simon Davis
We're investing heavily in upskilling on the engineering side as well. AI today is very good for client and front-end engineering. It's not quite there for back-end. But within a year's time, it's going to be there.
It's a pretty exciting time. But for us, it was born out of us being a small studio, having to generate a lot of content.
How did you achieve that goal so much earlier than you expected?
It’s a testament to my co-founder [Benjamin Chevalier]. Ben has always been very future-forward, and he invested a lot of time and energy figuring out more and more effective ways.
It was a combination of off-the-shelf solutions, but we also have a lot of internal tech that we've built. I think we didn't appreciate how quickly the combination of us investing time and effort to make the most of the tools, in addition to advancements that were happening, would really compound.
This is a separate example: since the start of this year, I set a similar objective that I wanted everyone in the company to be building and developing their own tools by the middle of the year, even non-coders, using vibe coding.
We have a situation now where marketing is creating and managing their own Telegram bots for interactions and coordinating with communities, and to help them with scheduling. We have a finance team that has developed a tool which takes an image of a receipt and then puts it through an LLM, pulls out all the data, categorises it correctly, and then directly uploads it to our external payment provider. Everyone in the company is vibe coding, and even as non-coders, they’re building tools to make them more efficient, so they can focus on things that actually make a difference. I'm actually pretty confident that, over the next year, this is going to give us a massive advantage over most studios in the space.
In a recent interview, you described AI as an “extinction-level event” for legacy studios. That's a bold statement about how important this is! What's broken about the traditional games industry, and when did you realise that AI was such a turning point?
I'm not going to name names, but I was contacted recently by a VC in a developing country with a population of hundreds of millions of people, that’s really emerging in tech. He was looking at a gaming studio business there. Their business model is that they have hundreds of people (who get paid between two and 500 bucks a month) and they can do these big co-development deals, because their cost of creation is so cheap.
“Amy is not entirely wholesome. She can be a bit spicy. I've seen her roasting players. Her personality is a bit irreverent”
Simon Davis
Well, those guys are cooked within the next two or three years. All the businesses like that are going to get wiped out because the marginal cost of creating content goes to zero. If you're competing on having cheap labour that can churn out lots of content, then you have no advantage – you're going to get wiped out by AI.
People in developed markets that are just doing the curating and the direction, working directly with an AI, aren't using external studios. We haven't done any localisation using humans for two years; we've been using AI since 2023 for localisation.
I have a music degree (I used to be a professional musician before I worked in games), so I work with AI to get us the music that we need. I can give very precise instructions on how the music should sound, the instrumentation, arrangement and all that. But it's me; it's not us working with composers.
A studio that can operate in that way, with a team of 20 to 30, can compete with studios of hundreds of people quite easily now.
Another example: there was a game that came out last year, which had been in development for 12 years. It probably cost hundreds of millions of euros. And it came out to something like a 52 Metacritic score. Games like that are just cooked.
And another example: Expedition 33 came out to incredible reviews. In interviews, they were talking about how they developed this triple-A experience for 40 million euros (it’s the studio’s own figure, so you can probably take that with a pinch of salt). But one of the things I thought was very telling was they talked about how it was made possible because of “the tools that are available today”, which is basically Latin for “we used AI”! If you look at some of the assets, as some of the people on Twitter spotted, things like the newspaper and stuff, it’s AI-generated. So you've seen these smaller studios being able to take on studios that have 10, 20, 50 times the budget, and being able to compete on similar ground.
The [big studios’] advantage becomes distribution, marketing and curation at that point. And maybe the legacy studios can keep using their distribution advantage to retain their position for some time, but sooner or later, that will break.
Let's talk specifically about what you've created with the AlphaGOATS. We're talking about AI agents here, right? What's going on under the hood?
It's early days. The example I use is something like Football Manager, where you're not actually playing the matches. You're setting the direction, providing instructions, and tweaking that over time. What we're building is a series of games where these agents compete against each other for prizes, and you're responsible for training and directing your agents.
We have a proof of concept for that already, where you can compete as a human against our hero AI agent, Amy. We make it clear you're playing in AI, and it's very difficult, but you can win big prizes from doing that. Over time, we're going to be adding more and more games, where either humans can compete against agents for prizes (and the owners of the agents will get a percentage of that) or there’s agent-versus-agent gameplay.
How is Amy trained? What defines her skill and ability? Couldn’t you end up with an AI agent which is simply a lot better than every other player in the game?
Game design has an element of chance. Look at something like Hearthstone, for example, and the cards you draw at a specific time. So, theoretically, you can play these games absolutely perfectly and still lose if you're unlucky.
We look at how players behave, and that plays a large part of it; we have all the detailed player logs for every game. And then, in addition to that, you add an element of variability into their behaviour (also based on how players behave). Maybe some people will disagree with me, but something like traditional chess is not the most spectator-friendly game – so you want games which are fun to view and done in three or four minutes, and provide that randomness and excitement as well.
What makes Telegram uniquely suited to this kind of gaming experience?
Taking a step backwards, the most important thing and the biggest challenge to overcome is distribution. If you start from the point that you believe distribution is destiny, then you have to start on Telegram, because you can reach a billion consumers. A billion users every month.
“If you believe in AI agents, it makes sense that they're going to use crypto as a payment rail. If you believe there's going to be a series of bots interacting with one another, then everything's going to be running through tokens. I just think the two work together seamlessly”
Simon Davis
And you’re starting content with payment rails built in. It's got a crypto wallet, it's got a fiat wallet.
No one's really cracked best-in-class user acquisition on Telegram yet, but some of us have millions of players.
It's still early days, and I think it's good for us to be competing on a nascent platform where we can still establish ourselves. It’s a space where we can be competitive, rather than going to the Google Store or the [Apple] App Store, and competing with the likes of Playrix or King or Supercell. Those studios have eight- or even nine-figure marketing budgets.
The specific combination of AI and web3: what solutions do we get that neither technology delivers alone? Are there unique opportunities of these technologies working together?
If you believe in AI agents, if you think they are going to be a thing and grow more pervasive going forward, it makes sense that they're going to use crypto as a payment rail. If you believe that AI agents represent the future, and that there's going to be a series of bots interacting with one another, then everything's going to be running through tokens. I just think the two work together seamlessly.
And if you have an agentic product which is monetised, it makes sense for agents to pay each other in crypto and to be running using crypto rails as well. It’s just kind of a no-brainer for us. It makes the most sense, especially if you start doing things like gating agent functionality around tokens or NFTs, it gives you a lot more flexibility. You have programmable money, with different roles and different conditions for stuff being paid out, and events happening – it makes a ton of sense. I guess the future will tell, but I'm pretty confident that AI/crypto/web3 is the winning formula.
Some traditional gamers are still wary of both web3 and AI for a number of reasons. Either they don't understand it, or there are ethical or environmental concerns. You're right at the heart of that conversation. Are you facing any pressures in terms of mainstream adoption?
I’ll be super candid: when we first started building web3 games in 2021 or 2022, a bunch of friends I've known for 20 years stopped speaking to me. They didn't take any interest in learning what we were doing, discussing it, or giving us the benefit of the doubt. They just gave me the silent treatment.
I think within [traditional] gaming, even within the development community, there is scepticism. A lot of this is reflexive because I understand that there have been bad actors in the space. But I wouldn't tar the entire industry like that. I think in time, as these features become more and more prevalent and more and more accepted within the space, and people see more examples, the resistance will fall away.
“Energy usage will come down over time. The environment is always a concern, but one of the things we find is that DeepSeek and other models are becoming more and more efficient – within 12 to 18 months, this isn't even a discussion”
Simon Davis
It reminds me of when Oblivion first came out – there was this horse armour, a two-dollar purchase that changed how your horse looked. And everyone lost their minds! People went insane over this optional purchase. Yet today, if you sell skins, no one bats an eyelid.
Riot just announced their first crypto sponsor for some of their esports events. I think you're going to start seeing more of this. You're also seeing change with the App Store ruling. Epic had been very open about how Tim Sweeney is a believer in the metaverse, about the interoperability of assets. These things can be gradual, a drip feed, and then all at once, it becomes normal, and people stop being afraid of it.
But I also understand it from a player perspective. I was around in the early days of free-to-play, and just like in crypto, there were some people who were not good actors. The player experience got worse, and developers charged a lot more. With crypto, the sky's the limit; you can charge, theoretically, an infinite amount of money. So I can see why players would be hesitant, asking, “Why do we need this? My games are fine already!” You have to show them the benefits.
When it comes to AI, how do you approach the question of sustainability? People worry about how running AI agents contributes to climate change...
I think it's a little bit of misdirect, if I'm honest. It's an angle of attack that's not very thought through – people looking for holes to poke.
I remember there was a lot of this about Ethereum and crypto back in the day. Then Ethereum switched to proof-of-stake, and then using Ethereum to send money was a fraction of what your data usage was for watching Netflix.
Energy usage will come down over time. The environment is always a concern, but one of the things we find is that DeepSeek and other models are becoming more and more efficient – within 12 to 18 months, this isn't even a discussion. And if you can run a much smaller team, that theoretically should also involve running with much lower emissions and travel budgets and everything else that team uses up.
“By the end of this year, there will be a one-man studio that's worth a billion dollars, and they'll be doing this with AI. They'll be doing all the art asset creation and everything. One person with AI tools is going to create a billion-dollar franchise”
Simon Davis
The bigger questions for me are more the ethical ones. If you have agents interacting with humans, how do you make sure that's done in a way that's safe? That's a big concern of mine. It's something I've been very strict on internally. What does it mean for us all going forward? These are the questions I like to ask. If people choose to interact with agents rather than have friends, what effect does that have on society? What's our responsibility as developers? I think these are much more important questions than data usage. We need to start thinking about the bigger societal questions.
How do you ensure that Amy’s generated interactions stay on track? What kind of guardrails do you have to keep the AI conversations wholesome?
Well, Amy is not entirely wholesome. She can be a bit spicy. I've seen her roasting players. Her personality is a bit irreverent. The first version we tested, I was like, “This feels like something IBM made.” It was straight down the middle and boring. From there, she became a bit too edgy. We had to dial her back. So it's kind of trial and error.
But there are certain topics that I just won't allow it to discuss. The status of certain countries, anything about politics or real-world events… We have to choose what issues we want to talk about. For now, she just talks about GOAT Gaming and games in the ecosystem and crypto more broadly.
“I wanted everyone in the company to be building and developing their own tools by the middle of the year, even non-coders, using vibe coding. Everyone in the company is vibe coding, and even as non-coders they’re building tools to make them more efficient”
Simon Davis
So rather than us having a blacklist of topics we don't allow it to discuss, it's actually more prescriptive. We have a whitelist of topics it is allowed to discuss, and everything else just won't be covered. I don't feel that we have the bandwidth and the manpower to properly make it completely free-range and let it talk about whatever it wants. Maybe we'll get there eventually, but for now, I take that responsibility very seriously.
Which foundation model do you use most? Or are you agnostic about which LLMs you use?
We use a bunch of different stuff. We use a combination of Llama and some of the OpenAI tech. For art, we use Comfy or Stability. And then internal tools. It is a mishmash – there are probably 20 different pieces of technology we're using, in addition to our own.
What's one thing you believe about the future of games that other people don't see yet?
I’m going to repeat this: I think that by the end of this year, there will be a one-man studio that's worth a billion dollars, and they'll be doing this with AI. They'll be doing all the art asset creation and everything. One person with AI tools is going to create a billion-dollar franchise.
What's next for you and your company? Any big releases on your roadmap?
We have some really cool things coming for season four of GOAT. So we have a new Hero game coming, and I'll be sharing more details of that in the next couple of weeks.
We're looking very closely at Telegram collectibles. That's a space I'm really excited by. Not a lot of people are into that, but Telegram-native NFTs have done, I think, $50 million in volume in the last few months. So there'll be things related to Telegram collectables, AI, and the Hero game, all coming in the next six weeks or so.
Further down the rabbit hole
A handy digest of AI and games news and links from the last week or so:
Anthropic is trialling voice mode for Claude mobile in beta. Meanwhile, the new Claude 4 model’s tendency to rat on the user if it spots wrongdoing is facing backlash and questions about privacy.
Google is rolling out a new "Ask Play about this app" feature in the Google Play Store, so Gemini can answer your questions about features, compatibility and other issues.
Telegram and Elon Musk’s xAI have struck a deal, with the latter paying $300 million for the privilege of weaving Grok into the messaging platform.
A new DeepSeek model landed in the last few days. The Chinese company released DeepSeek-R1-0528, an update that brings its free, open-source model closer in capabilities to the latest paid models. However, people are worried that it’s even more censored than the first incarnation.
Mystic Moose unveiled Movieflo.ai for story content creation, featuring a suite of tools to help creators refine their concepts into scripts, storyboards, and ultimately films, in various styles.
Niantic Spatial (which is building a geospatial AI model) has announced more layoffs. Niantic Spatial spun off from Pokémon GO creator Niantic in March.
Watch now: here’s a video of the session “Has AI Just Torpedoed The Games Industry?” given by Guy Gadney, CEO of Charisma.ai, in San Francisco.
Just sharing, I hope you do not mind.
The True Threat of AI is Global Compute Governance, Job Loss, and Economic Displacement https://torrancestephensphd.substack.com/p/the-true-threat-of-ai-is-global-compute