10 Months of Unity Dev with Claude Code transcript

My workflow is that I treat Claude code as a junior engineer. The reason why that works for me is because I have a very clear vision of what the game is going to be. All the design in what I expect for the user experience to be, I already know in my head. I've already planned it out. I know what it should look like on the screen, even before it's been coded. So, what that means is the only way I can verify if Claude has listened to what I asked is by QA testing.

The downside is that means that I spend almost all my time QA testing, not programming, not making art. 95% of my time is spent QA testing because of the workflow that I've chosen. My workflow is purely visual when it comes down to how I interface with the game. I have a very clear, specific target user experience. So, I'm kind of tapping into a lot of that visual library and trying to pull it in. I think that's very important if you ever use my workflow is because you have to protect your taste at all times.

Because I have a very clear vision of what I'm trying to build and how it should feel, anytime I notice Claude deviates from it, which is basically all the time, I QA test it and then give it feedback. In this case, QA testing is basically playing a broken game, studying it, and giving feedback. And that's one area that I'm looking into automating. I don't think it's worth the cost right now. I think I've heard this concept of computer use with Claude code, but I really don't want to use API cost to try to figure out some sort of tool that would help with the play testing a little bit.

Cuz right now, I think the plan that I'm on right now it's just monthly and it's $100 and it's very predictable cost. As I build more confidence on the workflow, then I I can increase that. But for now, for my needs, that plan has been enough for me to work on this 8 to 10 hours a day if I want to. All right. So, there's a new feature that I want to make in my game. There is this overlay that pops up when I try to purchase something that is not unlocked.

When I press enter, you'll see the not unlocked yet overlay that pops up. I really like the visuals of this. And one thing I wanted to do is add that into my shop. To to show you what the shop looks like right now is if I go into the game and I proceed to go to the shop. Right now, if I can't afford something, the item just jiggles. But I want to better communicate to the player of what's going on. That's the next task.

And this is going to be a good example of my workflow where it's just me talking. I don't really think about too much about the technical details, just the design intent. What is the user experience? Another thing that I want to work on is the animation that plays when the player collects something in the shop. Right now, when you buy something in the shop, it just disappears. Let me give you an example. If I come here on this thing that's for sale and I buy it, it disappears.

And that's not my preferred user experience. And what I want to add is a fly effect of the item being purchased to where it's directionally going. Hey Claude, I need you to create a new feature for me. Do you know the overlay that pops up in the unlock screen and the character screen that says not unlocked yet? I need that implemented into my shop. And it's important for you to do no guessing here. I need you to actually read the code.

So, that's number one. And then number two is the actual successful purchase of the item in the shop. And in this case, I want to use the fly animation system that we already have. I believe the feature that we first worked on for the fly effect animation was the gold coins. So, do please reference that for what is the correct pattern to implement. Yep, looks good. Hey Claude, can you go ahead and do all of the necessary wiring for me, please?

Let me know when it's ready for me to test out in game. Thanks. All right. So, it looks like it is ready. So, let's go ahead and go back into Unity. I'm going to try to attempt to purchase this thing that I cannot afford. Let's see what happens. Okay, I do not see the overlay. So, now this is the fun part is explaining to Claude that it's currently not working. Hey Claude, so I tried to purchase an artifact in the shop and I do see the jiggle where it tells me I cannot afford it, but I do not see the overlay in the shop.

I was wondering if you could help with that. And also Claude, I'm going to attempt to purchase an item and see if the floating or the travel animation is in. So, here's my attempt. I'm going to for this, I'm going to turn on the audio just to make sure that the audio is also wired up properly. Perfect. So, the floating animation is all in. It looks like it is probably working. Okay, that's the issue. I don't know why it had this hidden.

So, it is working. If I turn this on, the game object, and go back into the game and press on it, it works. So, it is working. It's just for some reason Claude deactivated the game the prefab. Hey Claude, so I need to make a correction. You did correctly implement this, but for some reason the selection indicator prefab had its visibility unchecked. So, it was turned off even though the overlay is actually working. Because the game object is hidden, I don't see it.

But if I recheck it so that it's enabled and then press the same artifact in the game, I do see the overlay. And let me just double-check the text for you. It says not enough gold, which is accurate. Usually I give Claude one or two chances for it to figure the issue out itself. If not, that's when I rely on logging. Perfect. It's working. Let's see what happens when I purchase a card. Perfect. It went into the card button.

Okay. So, this is the change list. Okay, looks fine. Committing is one of my favorite things to do and having AI do it for me, it's awesome because if something breaks in the future or if something's not working, I can tell AI, "Hey, can you go ahead and check my commit history and find out at what point what change broke this?" I'm going to say yes. And the work should be committed. I hear a lot of people talking about how they use a lot of instances of Claude code and they're managing a bunch of different Claude windows, Claude code windows, and I am I'm already overwhelmed with the amount of QA testing because I can't keep up with the code fast enough.

I think one thing that I've realized is that I really can't use gigantic workflows that are created by experienced engineers and programmers. They don't come from necessarily an art background. This is my guess. [clears throat] I feel like for me, my workflow works from for me because it's simple. I came at it from more of the visual arts side. I do look online for what other workflows are out there. And usually I find that anytime I see a workflow that is done by an engineer, like someone who was an engineer before LLMs or a programmer and they did it professionally, I have very hard time following because their workflow is very different than what is intuitive to me.

So, I think that's one of the things that I've been also trying to figure out as I try to learn more about what kind of workflows work well with my style of game development. And the really nice thing is about agentic coding, I get this question often about like what is my workflow? That feels like the wrong question to ask somebody. And I can only speak for Claude code right now. It's very capable to adapting to your workflow.

The hardest thing to maintain is a clear vision of what you're trying to create when you're using Claude code because at every step, there's deviation. Because I played a lot of games and I've studied art, I have very clear idea of what the game should look like. Now, because the models are so powerful, I can just chat back and forth. One of the chats that I had, I just kept it open for I think about a week, was to create this character selection screen.

At no point did I went into plan mode. For me, it feels like art. If you've ever taken like a drawing class, where it's gestural, where you're not trying to get the perfect line, you're not even trying to get the perfect sphere. I feel like sometimes that's how I'm programming, where I just it's really hard to polish something that's not there. So, in my first passes, I'll just try to get the feature in at a semi-functional state, and then I'll just QA test it, and give feedback to Claude until it feels right.

And for me, it feels like art because it's either very gestural, conceptual, and then also, there's this term that I haven't looked up, but when I was a college student, they call it master copy, where you would go to like an art museum, and you would sit down, and if there's a painting that you really like, you bring out your sketchbook, and you try to recreate it in the best way that you can. And the idea is that by copying a piece done by a master, that you can hopefully learn something new from it.

This isn't about creating a final product, it's about like learning. And I feel like I do that a lot in my workflow. If you're asking me, Daniel, like, how do I even get started with using Claude code for game development? Don't worry about other people's workflow. Install Claude code, have a vision for what you're trying to build, and just build it. I think you don't need to be a programmer or coder to get started. You will learn a lot about coding and programming by using the tool.

You don't actually have to code to learn it. And I think while it is true that I didn't really know much about coding a year ago, I can't say that at all anymore. And from there, find a section of a video game. I think UI is a really great place to start when it when it comes down to this. And really pay attention to how it's designed, what sounds does it make, how does it move, what is a user input, and try to recreate that in your own project.

And I'm only speaking to people who are not um programmers, coders, or they're interested in a workflow that is similar to mine. And also, for anybody that wants to actually see my workflow in full, I really want to be respectful of your time, and the reason why I haven't really gone too deep into my workflow, is that I want to first validate it by releasing a game on Steam that someone paid me for, that reaches some sort of success before I start showing my workflow on YouTube.

And my hope is that this game will be that. And if it is, I'll show my full workflow. But I I just don't like the idea of adding more noise on YouTube. Because a lot of this workflow that I'm using feels experimental, and I need some sort of validation before I start thinking that it is a net positive to actually share with people. Because if someone looks at it like, this looks like this looks really bad. What are you doing?

You have no idea. And I'll be like, yeah, that's that's how I feel, too. If there is a lot of criticism with how I work, I don't really I I feel those, too. But the reason why I don't stop And the week The reason why I keep doing this is because it's so much more fun than any workflow I've ever done in game production. And part of it is I don't have to be in meetings. I don't have to If I want a feature in, I just get it in.

And also, no one lies to me about it, whether or not something can be done or not. As an artist, when I'm working in game studios, especially when I was more junior, there'll be like these really cool things that I wanted to do, and I'd talk to somebody who's very technical, they're way smarter than I am, and they don't differentiate between something that can't be done versus something that they don't want to do. And often times, when it's presented to me, it's, oh, we can't do that.

But what What I learned later on in life is it usually means they don't want to do it. And the nice thing about working with Claude code is if I just want to build something, I just build it. I think a lot of this stuff is so much simpler to just build yourself with LLMs. Because in teams, it's really hard to find alignment, and to just agree on what is being even being built. And I can only speak for as an artist working in other game studios.

Most of the work that I've ever done has been thrown out, in the garbage. Someone paid me a lot of money to make it, and after 4 years, it doesn't get released, and no one will ever see it. I've worked on this really cool like battleship that I was so proud of. And I worked on it for 5 months. Keep in mind, this game, I've been working on it for 11 months. Now, imagine if I spent 5 months of that just working on one piece of art.

It's huge. It's just It has There's just so much going on, and I felt so much, oh, this is so cool. I I I can't wait until it's public knowledge, so it's not under NDA. And something about so much of that never getting released, the company, you know, going under, or they just decide that they don't want to go forward with that project, or whatever. Whatever the rationale is. I don't know the details of why some of my work doesn't get released.

I think there should be a bigger backlash from not the gamers, right? Gamers, I feel like they make sense to me. They're they're passionate, um sometimes maybe overly so. It's usually the people who are like in charge of the companies. I don't know what it is. Most of the time, they don't even play video games, they don't care about video games, they're just smart and they're good with numbers. Now, here's the thing is, I'm guessing here, because I've never been one of those people.

I can't even relate to it. I think that's the thing that LLMs make me very excited about. There is the hope, right? Because I still have to do the work. I'm working just like a regular job. I think I spend about um 8 to 10 hour every single day. Um but it's not hard work. I think this is one thing that people need to realize is that it's hard work when you're working at a regular job, where people are telling you what to do.

I have more agency on how I like to work and what I work on. Some days, I don't want to do art. So, all I do is I just work with Claude code, maybe even on my phone if I need to. But some days, I don't want to use any Claude code. So, what I do is I'll draw. I'll make a 3D model. I'll make a hat. Like, there's a hat. That was done in a day where I didn't feel like using Claude code. I actually I think that was a day where Claude code was down.

I was like, oh, like, I literally don't know how to develop this game without Claude code. It was like a terrible realization, where I was like, oh, wait a minute, if the servers are down, this dream is gone. Like, the dream of game making this game. I I I don't know what to how to feel about that. Hopefully, in a year or so, I'll have like a backup local model that doesn't go down. I can only speak for an artist. I've led a team of artists, and the art team feels like a little bit of a punching bag sometimes in game studios.

It's nice to not feel like that when I'm using Claude code. There's so much PTSD from like working as an artist in the game industry. So, when a tool like this comes, I mean, for me, it feels like magic. I still sometimes can't believe it actually works. It It is crazy. Because it's crazy if you knew what my skill set is. Because I know there are very talented developers out there that can do stuff like this without agentic coding.

But I think the reason why I'm blown away by it is because I know where my skill set is. And I do feel like um the game that I'm building, whatever you think about the art style or the quality, all that stuff, it's only within context of me knowing my skill sets that I'm like, wow, this is amazing. There's something that's really annoying. I I don't like it when I get interviewed by people who don't care about video games, so that I can make video games.

This is before LLMs. It's like going into an interview, and the person on the other side could care less about video games. Like, they don't care. They're just like, their thing is, I have this thing I need created. Are you going to do that for me? And for me, um I've done that a few times, and it just doesn't feel good afterwards. Even if you get the job, and you work for them, the the work the work experience is like that.

This is only based on my experience, and not all of my experiences have been this way. I just mean, it sucks when you're in that environment. This is my first solo game, but it's not Something about proving it to myself is more important, and proving my to myself that I'm capable of making a game that a good chunk of people enjoy playing.