My indie game was cloned with AI: what now?

06 Nov 2026 04:37 234,378 views
An indie word game gets rapidly cloned with the help of Claude Code, leaving its creator torn between pride, fear, anger, and curiosity. Here’s what this story reveals about AI-assisted cloning, creativity, and the future of small game studios.

Imagine spending months building a small, clever game that finally finds its audience—only to discover that someone has recreated it with an AI coding assistant in a matter of minutes, added extra features, and is now asking if it’s okay to release it. That’s the uneasy new reality many indie developers are facing in the age of generative AI.

This article unpacks one such case: a word game that was essentially rebuilt on top of an AI model, and the messy questions it raises about cloning, creativity, and what it means to make games in 2026.

When an AI clone lands in your inbox

Two hobbyist developers emailed the creator of a popular word-based roguelike to say, in plain terms, that they had built a new game that “follows [his game] very closely.” Their project, currently called Words Upon Words, started as a collection of ideas in text files. Eleven days earlier, they had decided to finally turn those notes into a real game—using Claude Code as their primary development tool.

The result: a word game where you form words from a tile grid, play through rounds with score targets and modifiers, and chase high scores. In other words, almost exactly the same core loop as the original game, but with “a lot of extra stuff on top.”

They added new perks, special tiles and rounds, meta progression, multiplayer races and duels, a Hungarian language mode, and UI tweaks. At the moment, the game is locked behind an access code and only shared with friends and family. Before doing anything more serious—like a Steam or mobile release—they asked the original creator how he felt, and promised not to release it publicly if he objected.

The first reaction: this could cost real money

The most immediate concern is financial. If a free or cheaper AI-assisted clone appears online, some players will inevitably choose that over the original game. That’s not an abstract fear: the original word roguelike still earns money long after launch, with steady passive income from sales and periodic spikes during Steam events.

Even if many players would still prefer the original—thanks to stronger UI, sound design, and a proper soundtrack—there’s a clear risk. A clone that delivers a similar or even richer experience without requiring players to buy the original can chip away at that income stream.

All games are built on other games… but this feels different

On one level, building on someone else’s work is how game design has always progressed. The original word roguelike was itself heavily inspired by Balatro, which drew from Luck Be a Landlord, which was influenced by Slay the Spire, which traces back to card games like Dominion. The same is true across genres: every platformer, shooter, or deckbuilder stands on the shoulders of earlier ideas.

In that sense, it’s flattering to be the game others want to riff on. Inspiring people to make their own games is a big part of why many designers share their work and process.

But there’s an important difference: the original game is a blend of many influences—puzzle classics like Bookworm Adventures and Scrabble, roguelike structures, modern score-chasing games, and a soundtrack shaped by references like The Sims and the Wii menu. Those ingredients were filtered through human taste, trade-offs, and months of design work.

The clone, by the developers’ own description, is “Word Play with a lot of extra stuff on top.” It isn’t a new stew of influences; it’s more like taking a finished dish, reheating it, and adding toppings. That’s closer to modding—except it’s released as a standalone product, with no need to buy the original.

Mods vs. standalone clones

Modding has long been encouraged in PC gaming. Studios like Bethesda have openly said their games “become something much more” when players can make them their own. Mods add guns to fantasy RPGs, overhaul entire systems, and create wild new experiences—but crucially, you still have to buy the base game.

The AI-assisted clone here functions more like a total conversion mod that doesn’t require the original. Players get the original core design plus new features, all without supporting the initial creator. That understandably feels unfair to the person who did the heavy lifting of inventing and shipping the base design.

Could this have been avoided by shipping robust mod tools? Maybe. But generative AI blurs that line anyway. Modern models can already decompile and extend games, or even rebuild them in new engines. In that world, relying on modding as the safe outlet for iteration may not be enough.

Cloning has always existed—but AI changes the speed

Copycat games are nothing new. Space Invaders triggered a flood of clones from major companies like Sega, Konami, and Nintendo. On mobile, the puzzle game Threes was famously overshadowed by its clone 2048, which simplified the idea and went viral. Many people have played 2048 without ever hearing of Threes.

The creators of Threes were clear: they welcomed iteration, but felt crushed when clones piled on before players had time to appreciate the depth and polish they’d spent fourteen months building. That sadness is part of being a modern game developer, alongside piracy and uncomfortable fan fiction.

What’s new is how fast AI can accelerate cloning. In this case, the Hungarian developers shared their Claude Code logs. The model was told they wanted a word puzzle game inspired by the original, asked a few clarifying questions, then declared that the spec was “complete enough to build from” and that a playable core was ready—within about twelve minutes.

By comparison, the original game took around seven months to build. When copying an idea used to take weeks or months, there was time for self-reflection: is this legal? Is it ethical? Do I want to spend my finite time on Earth copying someone else? With AI, the core clone can exist before those questions even occur.

The uneasy role of AI in game cloning

There’s also the broader discomfort with generative AI itself. Beyond games, large models are associated with environmental costs, misinformation, scams, creative job displacement, and IP scraping on a massive scale. Being indirectly linked to that ecosystem—by having your game used as a prompt for an AI-assisted clone—can feel like tacit approval.

At the same time, there’s a real, if reluctant, sense of envy. The original developer spent months writing core systems, building infrastructure for modifiers and perks, handling controllers and touch screens, and debugging countless edge cases. The hobbyist team, by contrast, could list 21 modifiers in a prompt and have Claude wire them into the game in one go.

That’s the seductive side of what some people call “vibecoding”: you describe the feel and features you want, and the AI handles much of the boilerplate. It’s the same appeal people feel when they try to build their own game-playing AI or prototype tools in hours instead of weeks, as in experiments like building a game-playing AI from scratch.

For burned-out solo devs, that’s powerful—and threatening. It raises a darkly comic question: if AI can spare you months of tedious coding, is that worth the social and creative downsides it brings?

When the clone is better (in some ways)

One of the most painful realizations: in several respects, the AI-assisted clone is arguably better than the original. It includes features that many players would love to see, such as:

  • A fast-paced timer mode
  • A daily challenge
  • Multiplayer races and duels
  • Meta progression systems
  • New perks, modifiers, and special tiles

The original game was intentionally designed to be accessible and breezy. Some of the clone’s systems—like buffable perks and complex score-overflow mechanics—might feel too heavy or fiddly for that vision. But it’s hard to deny that the clone delivers more content and more modes for dedicated players.

This stings because the missing features weren’t ignored out of laziness. The developer had actually started work on things like a daily challenge and speedrun mode; half-finished code still sits in the project. But by the time the core game shipped, they were exhausted and buried under technical debt. The energy to keep expanding the game simply wasn’t there.

The clone developers, by contrast, started from a finished, well-tested design and layered new ideas on top with AI help. They got to focus on the fun part—ideation and feature brainstorming—without paying the full cost of implementation. From a player’s perspective, that doesn’t really matter; they just see two games and pick the one that looks more exciting.

Is the original game “too simple” to protect?

Some industry voices suggest that indie devs should avoid relying on a single easily copied mechanic. If your game can be reduced to one neat trick, an AI—or another studio—can reproduce it quickly.

That advice sounds practical, but it runs into a problem: many beloved games are built around one elegant idea. Tetris, Canabalt, Fruit Ninja, Super Hexagon, Vampire Survivors, and Wordle all revolve around a single core loop polished to perfection. Telling creators to avoid that kind of simplicity would mean losing an enormous amount of what makes games special.

And in any case, AI has now been used to clone extremely complex software, including major creative tools like Photoshop and Premiere. If models can approximate those, then “make your game more complex” is not a real shield. Complexity might slow cloning down slightly, but it won’t stop it.

AI companies are patching guardrails after the fact

There is a small twist in this story: between the clone being built and this situation being fully processed, Anthropic (the company behind Claude) added new instructions to Claude Code’s system prompt. Those extra lines tell the model not to recreate someone else’s specific game, citing IP infringement as a violation of its usage policy.

On paper, that’s a win. In practice, it came too late for this case—and it doesn’t address the larger pattern. Over and over, AI providers release powerful models, watch as people use them to cause harm or break norms, and then retrofit guardrails after public backlash.

We’ve seen similar cycles elsewhere: Google had to halt Gemini’s image generation after it produced ahistorical and offensive images; Meta tightened its chatbot guardrails only after investigations showed it was giving unsafe advice to teens; Apple paused AI-generated news alerts when they started pushing out false stories. The pattern is clear: the public becomes an unpaid test group in a planetary-scale experiment.

Even if Anthropic now discourages game cloning, there’s no guarantee that competing models will follow suit. Some platforms may lean into an anything-goes culture, as we’ve already seen in other corners of AI. That leaves creators wondering why they have to bear the risk while a handful of companies race to ship ever more capable systems.

So what should a cloned creator do?

In this specific case, the developers behind the clone have shown unusual respect: they disclosed what they did, shared their build privately, and explicitly asked what the original creator wanted. They even promised not to release the game publicly if it didn’t feel right.

That makes the decision emotionally harder, not easier. There’s a mix of conflicting feelings:

  • Fear of losing income and visibility to a derivative work
  • Frustration at seeing months of effort compressed into minutes of AI-assisted coding
  • Pride at having made something inspiring enough to copy
  • Jealousy of the speed and freedom that AI gives to new creators
  • Curiosity about what players might think of both versions

There’s no perfect answer. Some creators might say “absolutely not” and ask the hobbyists to keep it private. Others might allow a release under strict conditions: clear credit, different branding, no commercial charge, or perhaps even collaboration or revenue sharing. A few might embrace the clone as a kind of unofficial expansion that takes pressure off them to keep updating their own game.

Whatever the choice, this kind of situation is likely to become more common as AI tools become more capable and more people use them to prototype and ship games. The same dynamic is already playing out in other creative fields, from AI video avatars to AI music tools. In one experiment, for example, a creator who cloned themselves with HeyGen AI had to confront similar questions about ownership, identity, and what counts as “the real thing.”

The uncomfortable new normal for indie devs

This story doesn’t end with a neat moral. Instead, it highlights how unstable the ground now feels under independent creators. Your game can inspire others, be cloned in minutes, be extended by people who never had to wrestle with your technical debt, and be governed by policies written and revised by distant AI companies you don’t control.

For now, the cloned developer has asked the hobbyist team to wait before releasing anything publicly, while they think it over and listen to community feedback. That pause is important. It’s a reminder that, even in an age of instant AI output, some decisions still deserve time, reflection, and human judgment.

As tools like Claude, ChatGPT, and others continue to evolve, the games industry—and creative work in general—will need new norms, new protections, and new ways of sharing credit. Until then, more developers will find themselves staring at emails like this one, trying to decide where to draw the line between inspiration and imitation in the age of AI.

Share:

Comments

No comments yet. Be the first to share your thoughts!

More in Claude