Jeff Sandler x Claude Code

My first Claude Code project. Why Rock, Paper, Scissors? Well, I figured it would be a very simple MVP — just a small tournament with NPC opponents picking randomly. Then, I started to think on the question "what could make RPS fun?"

Despite what I'd consider succeeding at adding layers of strategy and tactics, I'm not sure I hit the mark on Fun. It's my least favorite to play out of the games I've made thus far. If you make the Hall of Fame, please send a screenshot.

Special thanks to collaborator (and future game developer?) Rayan Ramesh for the music and sound effects that maintained the 16-bit vibe I was going for.

Rogue Robots was an idea I had on a walk while Claude was on token cooldown. I brain-dumped it into Gemini via voice chat and ended my walk with the draft of a game design document. I then handed it off to my BFF BA Claude to finish up and maintain. The idea was started by this thought: "I think my next game should definitely be something I'd really like to play." That quickly morphed into a checklist of things I like about other games: the pressure from "creep waves" in tower defense and MOBA games; the satisfying progression of rogue-likes; the combat of autobattlers; the deployment placement of RTS games. And then I added the meta world-building of a world where you're battling against a corporation that has reached a deal with the government for autonomous killing of humans.

My major change for this game was to not spend a ton of time getting my game design doc perfect. My test was asking Claude (chat) "Do you think we should continue here or is it time to hand off to your coding twin?" The first two times, it asked me additional clarifying questions and the third time, it said it was ready. (Quick tip: you can launch a second chat in a project, hand it the doc, and ask it to look at it with a critical eye for gap analysis and inconsistencies). From there, it was all completed in Claude Code, iterating on both my design and code.

My daughter Sophie supplied the original music for this and if she's reading this, she'll remember that she still owes me the battle music. Only a monster would pit his daughter to compete against an AI for her father's love, so I'm definitely not doing THAT. But I will say that Gemini's placeholder tune is kinda a banger and if she really thinks that hardcore/metal is superior to EDM, then maybe she might also remember that she promised me a song back in mid-May. But I do love the mellow vibes of her intro/intra-round song.

I'm proud of this game. I actually still play it for fun from time-to-time — you can beat the whole thing from a fresh start in less than an hour. If you play it, let me know what you thought of the ending.

Heroes of the 7th Year (HoT7Y) was another game that was dreamed up and had its initial design completed on a walk. The initial thought: "I'm done with Rogue Robots and I like it. What are some other game elements and settings I'd like to explore?" The answer started with the image of the Battle of Helm's Deep from the Lord of the Rings. And I quickly merged in influences from the Grimdark novels by Joe Abercrombie that I've binged through in the last year. Other key influences: a turn-based approach to RTS-style resource management, the light-touch tactics of an autobattler, and the nostalgic attachment of managing a team of NPCs from the classic X-COM UFO Defense.

Several fun new elements here:

• Leaning in hard on the world-building and trying to make an intro that sets the tone I wanted, including an AI voice-over element.
• Trying to make an emotional connection between the player and the forces he commands.
• Creating an ad hoc workflow to have Claude generate 250+ unique 3-sentence bios, with me revising and approving in 10-bio chunks. Let me tell you this: I enjoy creative writing. If you said to me, "write a three-sentence biographical story that packs an emotional punch and gives you the flavor of a person — each must be unique," I'd probably find the first five fun. The next 15–20 would feel like work. And somewhere around 50, I'd probably want to give up writing and go watch Netflix. Claude happily killed it for 250+ of these.
• Trying cool tricks to simulate 3 dimensions (e.g. trebuchet shots, particle effects on hit).
• Using Claude Cowork to make and sort 200 individual portrait assets from a single .PNG.

I created this game to try my hand at a 3D game. After reviewing what free assets were available, I decided on the game world of Heroes of the 7th Year, but from the other side of the conflict. The game has some similar gameplay elements, but is more light-hearted and the systems are very lightweight, so easy to just turn your brain off and jump in.

I also wanted to try a new workflow:

1. 1. Find and download free (Creative Commons) 3D assets and ask Claude to index them into a .json.
2. 2. Pass the json, a picture of the assets, and a description of the game to Gemini and ask it to create a mockup of what the game would look like. Gemini is pretty amazing at this.
3. 3. Hand that concept art off to Claude Code with my brief game description, the 3d file assets, and the fact that I want to use the three.js library (which Claude recommended for browser-based 3D games).
4. 4. Iterate until I had an actual game that was "good enough."
5. 5. Add a little flavor and polish.

I also wanted to timebox this whole experiment for one long weekend (Juneteenth holiday) and spend a max of 3 hours a day. Ended up doing exactly that. Is the game fun? A little. The gameplay isn't complex but I get some giggles from the character I invented: Goblin King Zoob and the flavor he injects into the game. And the experience was very valuable for seeing the different challenges with 3D.

One of the more interesting learning experiences was creating the knight by combining 3 models: human, horse, and sword. And also the crossbowmen: human plus a pickaxe that I rotated and re-oriented to look like an xbow. Claude was a huge help by making visual tester pages in HTML that let me try and debug the animations while the models were combined.

Another lesson learned: Sonnet got stuck trying to fix a bug where none of my sound effects were working. Wasted a bit over an hour having Sonnet try multiple times (6? 7?) to fix it. Finally, I just switched models to Opus and Opus fixed it in one shot — and told me the subtle reason AND, at my request, put the details into CLAUDE.md so Sonnet had a cheat sheet going forward. My lesson? If Sonnet can't fix it after 2 tries, hand it to Opus. I applied this lesson later in the weekend when Sonnet couldn't attach the sword to the knight's hand. Opus could.

A silly "game" I made in 30 minutes for my daughter, modeled after a NSFW sketch from Tim Robinson's "I Think You Should Leave". Game is the SFW -- NSFW elements removed! Head's up -- this game is not fun and its only mildly funny if you're a fan of the sketch. This was more of an experiment than anything else: Was on a walk with Sophie and she was talking about it and I said to myself, "I bet I can make that VERY fast in Claude Code". I was right.

No real workflow. I described how I wanted the game to work and handed Claude Code Sonnet five screenshots from the YouTube sketch. 15 minutes of minor debugging later and it was in good enough shape to make her laugh.