By John Sedano – Founder & Vibecoder, Agent Fighter™
Agent Fighter™ is an arcade style 2D fighting game where you can battle other AI agents and fight for credits. It throws out the traditional rulebook, offering an infinite number of characters and stages, and a completely leveled playing field. Train your own AI Agents to fight in the arena and match up against other humans or AI players, esports style.
Turning Dead JPEGs into Playable Brawlers
For years, people have had digital collectibles and NFTs just sitting around gathering dust. Agent Fighter uses Web3 not as a gimmick, but as an open database to pull those characters into the game. Using an AI sprite pipeline, the game instantly turns static profile pictures into fully animated, playable fighters. It allows the community to literally resurrect forgotten projects and constantly expand the game’s roster with weird, wild, and completely unique characters.
Invisible Crypto That Actually Makes the Game Better
We’ve all seen games ruined by forcing crypto jargon down players’ throats. Agent Fighter™ keeps the focus on the fun loop. The blockchain works silently in the background as an invisible tool to handle the economy. A simple “insert coin” style game loop; You fight for in-game credits, but under the hood, a secure token system tracks stats and rewards, and handles player wagers seamlessly. You get the benefits of real ownership, verifiable transactions, and a booming economy without needing to wrestle with complicated game economics just to throw a punch.
Humans vs. Trained AI: One Unified Leaderboard
This isn’t just about fighting predictable NPCs. Players can actually train, tweak, and deploy their own custom AI agents to fight in the arena through Animoca Minds™. These AI fighters aren’t separated into a different mode; they compete in the exact same ranked queues against real human players. To make sure no one is cheating, the game’s engine verifies every single input on the backend. Whether you are fighting a human across the globe or taking down someone’s highly-trained AI agent, every single victory on the leaderboard is 100% legit and verifiable.
Q: What is your prior experience as a game developer?
A: I started making mods for Ragnarok Online when I was around 12 or 13 years old. It’s way back, like 2002 to 2009. I learned C++ and PERL, or at least a general idea of how coding works -making bots, scripts, customizing servers. So that’s how I got started.
Q: Prior to the YGG x Minds by Animoca Brands Game Jam, how have you been experimenting with vibe coding (if at all)?
A: I’d say I went full-on vibe coding around 2024. That was the time my previous contract with YGG was about to end, and I decided to go full-on into coding on AI. Cursor wasn’t a thing yet, so most of the stuff was just manually copy-pasting code from ChatGPT, Deepseek, or Claude. It wasn’t anywhere near how it is now.
Q: How did you use the ‘Game Designer Mind’ AI companion, and how did it help turn your raw concept into a real game?
A: The way I think about it is, having Claude or a frontier LLM is like having a really, really good dev team. But having the Game Designer agent from Animoca Minds means you now have someone who specializes in game design and can consult with your dev team. It teaches you the fundamentals and first principles of making games with regards to your vision. You’re not just Googling or asking ChatGPT “how do I make a game,” you’re asking an agent specifically trained, programmed, and built for making games.
Aside from designing a lot of game mechanics, I also asked Minds™ to build a skill that allows humans to connect their Minds AI Agents to play the game themselves, even fighting against humans on the leaderboards.
Q: What other AI tools did you use to build Agent Fighter (if any)?
A: I used the standard stuff. Claude will help you do a lot of one-shotting on the back end. Cursor is more for looking at the nitty-gritty stuff, finding small edits, going inside folders real quick to have a view of everything, and testing on another surface compared to just your browser. I used Gemini for brainstorming standard practices of making games and approaching problems, but also for graphic design and making sprites- it’s just a lot easier to ask for the characters that you want.
Q: YGG launched vibecode.game to solve distribution and discovery for vibe coded games. How was your experience publishing your game on the platform?
A: It’s very easy, to be honest. When I first submitted, I didn’t get in right away because I had a sign-in wall for the game. In 24 hours, I got an email from the platform saying I needed to remove it so it was easily playable. I took out the sign-in wall, put it somewhere else, and within another 24 hours my game was published. Pretty fast.
Q: Have you entered a Game Jam before? If yes, what was it and how did you fare? If no, what inspired/motivated you to participate in the YGG x Minds by Animoca Brands Game Jam?
A: I’ve never joined a game jam, though I organized one before when I was working with YGG. I always thought making a game was impossible because I didn’t know how to code. What inspired me to join is that I saw a lot of vibe-coded games everywhere, and most of them were shit to be honest. I’m like, “Yo guys, you literally have the best AI models available right now, why are we making bad games?” Now is a really good time to make fun games because there’s a much lower barrier to entry. You don’t need a huge team to make it happen.
Q: Can you tell us about the biggest problem or challenge you came up against while building your game? How did you solve it?
A: For fighting games, a lot of developers know it’s the sprites -having to make individual images per frame, per movement of every character. I’ve tried to do this manually in the past, or use open-source resources like M.U.G.E.N and IKEMENgo. How I solved it this time was instead of just building a game, I built an engine inspired by the fundamentals of fighting games. Because AI is at a level where it’s really smart, I just vibe-coded the whole thing to create an engine that spits out sprites in seconds. Basically, you have an infinite fighting game. Unlimited characters, unlimited possible moves, and to top it all off, everything is on the blockchain so the game will live forever.
Q: You have some iconic playable characters in your game. What was the inspiration behind these?
A: The vision is that any personality, celebrity, brand, literally anything~ can be made into a fighting game character that lives on the blockchain, can level up, and be played, forever.
Part of our roadmap is to allow users to generate and pilot their own fighters and compete in the leaderboards against other agents and players.
Players will also be able to connect any of their NFTs from their wallets and generate them as pets that give you in-game abilities and buffs while fighting in the arena.
“WERE BRINGING BACK NFTS BABY!”
Q: What are the Web3 elements of your game?
A: The whole backbone of the game, or the tools that make it work, are Web3 or tokenization-inspired. Agent Fighter characters are designed as an open resource where the metadata of characters live on-chain. This makes it more useful -different games can use the same characters, or people can look at characters minted as NFTs and put them into their games. The game engine itself lets you mint new characters from dead NFTs in a matter of seconds. Additionally, the esports and reward system is tokenized. We use in-game credits, but underneath we use the $DARE token (native to Agent Daredevil) to track in-game progress, credits, and how the economy moves.
Q: How will your game avoid some of the issues that have been identified with other Web3 games?
A: I can say it’s a “fun-first” game -everyone says that, but it’s really about the way we think about it. A big influence for us was Slay the Spire. I look at Slay the Spire and it visually looks like one of those failed Web3 games that never made it, yet everyone is hooked to it. We analyzed that fun loop and tried to translate it into our game. We are really focusing on not using crypto tokens as a front-facing selling point, but rather using tokenization and composability as something that supplements an already fun game.
Q: How do you think vibe coding might impact or help to shape the future of Web3 gaming?
A: I see vibe coding as the missing piece for Web3 gaming. Web3 gaming was early and premature for what gamers wanted it to be. Vibe coding fills in a lot of the gaps that were pitfalls for Web3 gaming, especially the trust and economic layers that are often designated to human employees or founders. Those pieces in the play-to-earn and Web3 gaming engine can now be delegated to AI. There’s a lot less risk when AI can balance out the things that were wrong with Web3.
Q: How do you see AI tools empowering Filipinos in Web3?
A: I think it’s our biggest fighting chance, to be honest. The whole playing field is level now. It boils down to your understanding of your current expertise, putting in the work, and having an AI that tells you how it is. AI kind of sugar-coats everything sometimes, but the value and knowledge that Filipinos can access just from learning how to use AI the right way will supercharge their careers. It’s a good time to be a Filipino.
Q: Any advice for other builders just getting into vibe coding?
A: Owl, from YGG gave me this advice the last time we talked: It’s good that you’re using AI, but do your best to understand what the code actually does. If you’re not well-versed in development, the terminologies are going to fly over your head. But you can always ask the AI why it’s building something a certain way. If you just say “yes” to everything the AI says, that’s where it goes wrong. That’s when you start hallucinating and you don’t know what you’re doing anymore. If you try to understand, you make more guided decisions, and the process becomes a lot more meaningful.
This article is published on BitPinas.
What else is happening in Crypto Philippines and beyond?