One-Person Studio with AI: MAXX Apps Founder on Building with Replit and Claude

Hello! We hear a lot about the one-person studio, where AI does the engineering and a lone founder does the rest. This week in AI Gamechangers, we meet someone who is living that life.

Spencer Wade is a designer by trade, with a long career designing interfaces and companion apps for games companies. Since April, he’s been building his own suite of apps for Roblox and World of Warcraft players, without a development team. He tells us which tools he uses, why the AI keeps insisting broken things are “done”, and why a human eye still matters most.

If a colleague forwarded this to you, welcome aboard. AI Gamechangers is a weekly conversation with games industry insiders putting AI to practical use. There’s a free CEO interview archive to browse, and a round-up of news and links at the foot of every edition (scroll down!).

Spencer Wade, MAXX Apps

Meet Spencer Wade, founder of MAXX Apps, a suite of gaming companion apps built on the Overwolf platform.

Spencer has been designing for gaming communities for over 25 years, starting with forum signatures for his clan and going on to work with Pipeworks Studios, Atari and Xbox. He spent 14 years as a designer at MOBA Network, where he worked on League of Legends companion app Porofessor.

In this interview, we explore how he builds apps like MaxxBlox and MaxxMMO using Replit and Claude, where the tools fall short, and his advice for other designers.

Top takeaways from this conversation:

1. Spencer builds with Replit, using Claude separately for ideation and build specs. Together, the tools have enabled a solo designer to ship three apps.

2. His advice to other designers: do your homework on what data and APIs are available before you start, and keep the AI in plan mode while you’re asking questions. Remember to personally test everything!

3. He’s happy as a one-person operation, but draws a line at AI generating core elements that players engage with constantly, such as game maps.

AI Gamechangers: You’re not an AI engineer. You’re a designer and UX expert. In your own words, give us your origin story. How did you get here?

Spencer Wade: I’ve obviously always played games, and I started my career in different industries. What happened was that I was playing games on teams – back then they were called clans – and every clan I was on wanted forum signatures. If you remember those, you’d have the animated forum signatures, and you had to keep them under a certain size.

I’d make forum signatures for all my team members, then that moved on to making logos for our clans, and then I started doing websites for our clans. I’ve been using Photoshop since around 1994, so it just evolved from there.

I went through different career changes and then went full-time freelance, designing websites and logos, mainly for esports teams. Then I started working contractually for Pipeworks Studios. I did websites for games like RollerCoaster Tycoon and Prominence Poker, working with companies like Atari and Xbox.

“If I’ve been able to increase somebody’s happiness through video games, that’s what I like. The more people I can do that for, the better, because games have been a big part of my life. It’s not just about tons of revenue”
Spencer Wade

Then I got on with MOBA Network about 15 years ago. I was the designer, and eventually became the design director, if you want to call it that. I’m the old one, so I’m the senior one!

I’ve always been in the gaming industry and designing for UI and UX, but not really in a traditional position. It was more general design. I was a jack of all trades rather than a master of one specific field.

Then I got here because I had ideas for apps that could be on the Overwolf platform, but I didn’t have the development knowledge to build them. I had the ideas and the designs in my head, and the world and the technology evolved to the point where I could actually start implementing them.

AI Gamechangers: What was the first thing this new generation of AI technology allowed you to build that you couldn’t have built before?

Spencer Wade: The first thing I did was pretty cheesy: I built a logo generator.

I used to have this website called Esports Logos, which was a marketplace where I sold logos to teams and streamers. That kind of got replaced by AI because the age range of the clientele was probably 13 to 18 years old, and they were basically like, “Sure, I’ll just do all this myself.”

So I thought, well, then I’m going to take advantage of it. I built a logo generator with pre-built and adjusted prompts that geared the AI model towards creating more esports-related stuff.

Then I did something else I was familiar with, which was small business. I built a SaaS called Accelerate for businesses wanting to grow their five-star Google reviews and things like that.

After that, I thought, “Well, I can kind of see how this works. I think I finally understand it enough to have AI do the development for Overwolf platform apps.”

The first one I built was MaxxDash. That was really because I wanted something that looked cool and that streamers and other people would enjoy using alongside streaming. It hasn’t really done that much, but that was me getting my feet wet. It’s focused more on live-streaming controls and things that streamers and casual gamers might use on a second monitor or as in-game overlays.

AI Gamechangers: What else have you built since then?

Spencer Wade: With the help of Overwolf’s marketing, direction and DevRel management, they said, “Why don’t you do something in Roblox? It’s our number-one game.”

So I did MaxxBlox! It’s a Roblox companion app. It tracks the games you play and your game time, and gives you a historical view and leaderboards for everybody who uses the app. It ranks you based on your play time per game, and it has quirky little ranks. When you start out, you’re an iPad Kid, then you’re a Noob, and eventually you graduate up to being an OG and then an Alpha Prime OG.

It has different leaderboards, and it also has Rivals integration. Overwolf provides the game events – kills, deaths, assists – so MaxxBlox can rank players based on their KDA.

It also offers in-game and second-screen overlays, including a sniper trainer, which is basically a reaction-time test focused on Roblox snipers.

Then the other one is MaxxMMO. That has fewer daily users right now than MaxxBlox, but people engage with it more. That’s focused on World of Warcraft. The main focus is Midnight, the retail game, and it gives users different tools. It tracks your characters, so you can keep track of your main and all your alts. You can scout other players, copy their builds and keep track of Mythic+ scoring.

It also keeps track of auction-house data across the realms, so you can look up items, set pricing alerts and receive alerts in-game when you can buy or sell something at the price you want.

“That whole mindset around ‘early access’ bugs me. I think if you’re releasing something, it should be there. There shouldn’t be early access over a period of years! My issue with a lot of the industry is just the release of unpolished things”
Spencer Wade

I’ve also just released the Classic Burning Crusade capabilities for the app. It’s basically becoming a portal where you go in and choose your game. It has similar features, but it also expands on those with things like leaderboards and full loot tables for bosses and raids, from vanilla WoW through to The Burning Crusade. There are in-game overlays as well, so you can check loot while you’re raiding.

And then World of Warcraft: Forever is being released around November 4, and right out of the box MaxxMMO will have a portal supporting Forever.

AI Gamechangers: How did you actually build them? Which AI tools and models did you reach for, and where did the “vibe coding” side enter the process?

Spencer Wade: I think I was on Facebook (because I’m older!) and I saw an ad for Replit. I thought, “Wow.” Clicked on it, tried it out, and started messing with it.

I use Replit, and Replit spins everything up for you. It can do the local API servers, the database and other elements, and it integrates different models and agents.

I also use Claude separately for ideation and working on the build specs. Claude knows the capabilities of Replit, so I work through the build specs with Claude, send those over to Replit and tell it what I want it to do.

That’s basically the technology. It’s all done in Replit.

AI Gamechangers: What have you found that the tools still can’t do? Were there moments when your own human experience was what saved the day?

Spencer Wade: Obviously, because everything is built on the Overwolf client, the AI can’t test things. That’s a big limitation.

It will say it’s fixed. It’ll say, “Done”. It’ll say everything is hunky-dory… and then actually it doesn’t work at all.

That wasn’t too much of an aggravation because I knew it going in. The main pain point for me is the design-related stuff.

I tell it, “This is how I want it designed. Here’s how I already designed it for you. I need it to do this, and I need it to look like this.”

It keeps telling me, “Okay, everything is done,” and then I look at it and think: this is completely horrible.

Even though Replit is using Claude and has design capabilities, it’s still limited. Whatever it has learned about design systems, it tends to keep reusing that. It gives me its version of the AI junk, and I have to sit there and fight it.

That’s one of the biggest aggravations, but it’s also where the human comes in – the human eye. I’m big on user interface and user experience, and it has no idea how a human actually interacts with something.

AI Gamechangers: How well did the apps integrate with Overwolf itself and the data available through the platform?

Spencer Wade: It wasn’t seamless, but it was pretty smooth.

Right around the time I started creating the Overwolf apps, they had documentation and resources available that Replit and Claude could understand. So it was able to understand how a native Overwolf app works, how the client works and also Electron.

There’s tons of documentation, but the AI would still get hung up on things. I’d have to go through all the documentation and tell it, “You need to store this in your memory. You have to keep this in the project specs.”

AI wants to do its own thing, so you have to reiterate these details over and over again.

I was updating MaxxMMO over the weekend and I was saying, “We’ve been working on this for months and you’re going back to the drawing board. I’ve already saved everything in the project scope and in your memory. Why are you doing this?”

I sit there and get mad at it. I’m like an old grandpa yelling at AI: “Why do I have to keep telling you this?”

It’s almost like having a teenager!

But in terms of communicating with Overwolf, there weren’t any huge hiccups or breakdowns. It was really about laying the groundwork for how everything interacts, based on the documentation and the APIs available.

AI Gamechangers: There’s been a lot of hype around the idea that AI could enable a single person to operate an entire company. From where you sit, does being a one-person operation with AI as your development partner actually feel viable?

Spencer Wade: I don’t know about the idea of a billion-dollar company. To me, if you put it on the scale of a billion dollars in revenue, you’d have to have more people.

But being a one-person operation with AI? I don’t have any problems with it. I think it’s great.

“Replit spins everything up for you. It can do the local API servers, the database and other elements, and it integrates different models and agents. Claude knows the capabilities of Replit, so I work through the build specs with Claude”
Spencer Wade

Originally, over previous years, I had a real hang-up with AI. I hated it. I was thinking, “It’s replacing people’s jobs”, and things like that. And that’s kind of what happened a little bit to me. But now, sitting here with it, I think it’s cool. I don’t have any big hang-ups about it, and I don’t have any problems working with it.

There are some days when I don’t look forward to sitting here chatting with the old AI, but I think it’s great that I’m able to do this and provide people with a product that I wouldn’t have been able to provide before.

AI Gamechangers: You’ve said that the MAXX Apps are about making games enjoyable. If we speak again in a year, what would success look like to you?

Spencer Wade: I guess it would be being where I’m at now, but on a bigger scale. I don’t mean having five or ten apps this time next year. I mean having more people using the apps. I’ve always enjoyed the community aspect of video games and connecting with people on that level.

I have a Discord, and I like opening it up and seeing all these people on there.

It’s cool because when you’re building apps for different games, you’ve got different genres and different ages of people. I’m sitting on the Discord for MAXX Apps, and there are people from all walks of life, interested in all these different things, and they’re interacting.

When you get to around my age, you have this nostalgia for enjoying games with your friends and other people, and it’s nice to see that.

So a year from now, it would be nice just to see the MAXX community grow and the user base grow, to be able to implement people’s feedback, their wants and desires, to make their games more enjoyable. I’m now actually able to provide that.

If, in a year, I’ve been able to increase somebody’s happiness through video games, that’s what I like. The more people I can do that for, the better, because games have been a big part of my life.

It sounds kind of nerdy, but video games were a big part of my life growing up, and they still are. It’s not just about tons of revenue – although revenue is part of it because I need income!

AI Gamechangers: Having now built several real apps with Replit and Claude, what would you tell another designer who has an app idea but no traditional development background? Where should they start, and what mistake should they avoid?

Spencer Wade: The first thing is that you have to do your homework on the data.

Whatever you’re building, you need to understand what’s possible and what’s not.

As a designer, you know how you want something to look, and you know what you want out of it. But you can’t just go in and say, “Hey, AI program, this is the design, and this is what I want. Just build it.”

I think that was my problem coming into it. I thought you just handed it over and said, “This is what I need. I need you to build it,” and the AI would just build it.

You actually have to put a lot of thought into what it would take to build it, what is feasible and what isn’t.

You have to teach yourself a little bit.

“There are some days when I don’t look forward to sitting here chatting with AI, but I think it’s great that I’m able to do this and provide people with a product that I wouldn’t have been able to provide before”
Spencer Wade

If you want to build something on Overwolf, for example, and it’s for a game, you have to know what APIs and what data are available. You need to know that going in.

That was a big hang-up for me. I thought I could just say, “Do this,” and it would do it, it would work, and it would come out how I thought it should.

The other thing I would do when starting out is ask the AI: “Can you do this?” And make sure it’s in plan mode when you’re asking questions and planning. If you don’t have it in plan mode, it can take what you’re saying as, “This is what I need you to do,” and it’ll just start going off into its own AI world because it wants to take care of the task and be done with it.

AI Gamechangers: You’ve worked in games for a long time. How does the industry feel to you today? What are the challenges and opportunities, particularly around AI?

Spencer Wade: I enjoy the games industry. I don’t have a lot of negatives about the current industry.

The main thing is that there are a lot of games released that feel unpolished, or stay in early access with issues forever.

Valheim just released 1.0, and it had been in early access for years. That whole mindset around “early access” bugs me. I don’t like it. I think if you’re releasing something, it should be there. There shouldn’t be early access over a period of years!

Maybe my issue with a lot of the industry is just the release of unpolished things, especially AAA providers releasing stuff that shouldn’t be released. Companies take all this time and then release things that shouldn’t be released in that state.

But things like microtransactions? I don’t care. I enjoy them. I really like loot crates and stuff.

“I had ideas for apps, but I didn’t have the development knowledge to build them. I had the ideas and the designs in my head, and the world and the technology evolved to the point where I could actually start implementing them”
Spencer Wade

What does bug me is maps that are kind of AI-generated. There are parts of games where the players are engaging with something constantly, and that is human engagement. I don’t think AI should be in those areas.

There are certain things I don’t like about where the games industry is going, and it’s when these huge user-interaction features are developed and ideated by AI. I don’t think that’s good.

AI Gamechangers: Finally, what are you building now? Are you concentrating on those three apps, or have you got the bug now and there’s something new on your workbench already?

Spencer Wade: I’ve been playing WoW: Forever. We got into the beta, and I’m getting MaxxMMO ready for that.

But the other idea I’d like to do next is another Overwolf app. It’ll focus on the main FPS games – Valorant, Apex Legends, Rainbow Six Siege. I’m going to throw The Finals on there because I like The Finals.

Basically, I’ll use the games where Overwolf has Game Events Provider support. It’ll use Overwolf’s game events to track players’ kills, deaths, assists, KDA, total kills, total deaths and all that. It’ll track their performance across every FPS game.

Then it’ll give them detailed timeline breakdowns showing every time they play and what time of day it was. I’m trying to tie in what their ping was when they had a kill or a death, along with their FPS.

It’s going to be a nerdy tracking thing, basically, so you can see how your performance changes at different times of the day and in different situations, including your network connection, your frame rate and your PC performance.

It’ll also rank every single user of the app. They’ll get different tiers and different ranks based on their first-person shooter performance.

Then, within each individual game, it’ll rank everybody who uses the app based on their KDA and time played. A lot of people like seeing how they perform and whether, historically, they’ve got better at a particular game.

The good thing about doing it with FPS games is that Overwolf can provide those game events, so I can track everything.

It’ll have a lot more graphical and animated badges and achievements as well. It’ll be more graphical and have a cooler design. That’s the next thing coming up.

I’m also going to update MaxxBlox with a new user interface and a slightly different design.

When you open the app, it’ll give you a rundown of everything new: the highest movers who have ranked up, featured players, featured Roblox games.

The app already keeps a list of all the top games being played, so it’ll show the highest movers and the trending games. It’ll become more of an overarching update system for Roblox.

And it’ll be more graphics-heavy, to appeal to the bold graphical style that Roblox users like.

Subscribe! It’s free, and you’ll receive interviews with industry insiders and studio leaders in your inbox. We have many conversations in the pipeline, including Brad Wardell, CEO of pioneering software developer and publisher Stardock, who’s working on Clairvoyance AI.

Further down the rabbit hole

Some useful news, views and links to keep you going until next time:

• MTG wants to double in size within five years, and CEO Maria Redin told PocketGamer.biz that AI means a team that made one game a year ago can now make three.

• Capcom is overhauling its RE Engine in a project called REX, aiming for a “game engine of the AI generation” that speeds up workflows, though AI-generated assets still appear to be off the table.

• Meta has unveiled Horizon Create and Horizon Studio, a mobile app and a browser-based tool that turn prompts into 2D or 3D mobile games playable in Facebook and Instagram feeds.

• It’s been on shelves for a month now – do you have your copy of Myth-Made Machines: The Stories Behind Artificial Intelligence yet?

• “Most of what gets called ‘creative control’ in GenAI game tooling right now is theatre,” says Simon Davis, whose former Mighty Bear team has left games behind entirely to tackle AI’s brand consistency problem as SecretSauce Labs.

• ZeniMax’s lawyers have shut down Fallout: New York, a browser game one creator built by directing Claude and a team of AI agents to make something “exactly like Fallout”.

• Choosing an image model for your art pipeline? Artificial Analysis has a handy comparison of quality, speed and price across 167 of them, ranked by blind human preference votes.

• The Pope, Leo XIV himself, has laid into AI slop, pledging on X that “the Church wishes to renew an alliance with artists and cultural institutions to safeguard our humanity”.