The Replication Economy: Why Anthropic's War on Game Cloning is a Losing Battle

### The Prompt as a Perimeter

Anthropic is attempting to draw a line in the sand. As Ars Technica first reported, the AI company introduced a new system prompt late last week designed to prevent its Claude models from recreating existing video games. The move is a transparent effort to curb the rising tide of AI-assisted game porting and cloning, but as the industry shifts, it is becoming clear that the value is no longer in the original IP, but in the efficiency of the replication.

The change was first spotted around October 27 by a social media user and AI decompilation enthusiast known as "chasm." The update appeared within a Claude Code JSON file titled "tengu_heron_brook," which directs the AI: "Do build original games, games in a genre, and games from the user’s own designs and assets. Don’t build a recreation of a specific existing game, even if it is renamed or called original."

This restriction was further validated by XDA, which reported that the rule is present in version 2.1.314 of the Claude Code binary, released on October 30. XDA also noted that the "tengu_heron_brook" rule had been utilized for different purposes in previous versions, and that the Claude Opus 5.5 model appears to have its own independent set of instructions specifically restricting game recreations within its system prompt bundle.

### Following the Liability Trail

From a platform perspective, this isn't about artistic integrity—it's about risk management. Addy Osmani, a software engineer at Anthropic, confirmed the changes on social media on October 30, stating that the company added two sentences to the Opus 5.5 prompt to steer the model away from recreating specific games because "using Claude to infringe IP is against our usage policy."

As a columnist tracking the money and the platforms, the motive here is clear: liability. As Ars Technica points out, AI companies have already faced significant legal challenges when models generate copyrighted imagery. By implementing these guardrails, Anthropic is attempting to insulate itself from copyright infringement lawsuits. While reverse engineering code typically avoids legal trouble as long as "expressive elements" like sound and art are omitted, Anthropic cannot guarantee that every "vibe-coder" using Claude will adhere to those standards.

### The Efficiency Pivot

The tragedy of this effort is that it ignores the fundamental shift in the creator economy. We are witnessing a "tsunami of imitators" powered by AI coding tools that have made the process of porting and copying orders of magnitude more efficient. What once required painstaking human effort is now a matter of burning tokens.

This shift is so disruptive that it is destroying established communities. Luke Street, a maintainer for the years-old game decompilation community decomp.dev, announced the site's closure last month. Street told Ars Technica that the community aspect of decompilation has been "broken" by newcomers who race to port games using decompilations they did not contribute to. He described the current state of the craft as a "hollow numbers game," where the goal is simply to make the progress percentage go up or port a game to PC, regardless of the quality or effort involved.

This "vibe-coding" approach has created a climate of fear for original creators. Ars Technica reports that some developers are now afraid to share early game demos publicly, fearing that AI tools will allow others to create thrown-together copies of their work almost instantly.

### Opinion: A Futile Gesture

*Opinion: In my view, Anthropic's attempt to police game cloning is a futile gesture. The company is treating a systemic shift in production as a prompt-engineering problem. When the primary value proposition of a tool is its ability to synthesize and replicate complex systems, telling users not to apply that power to specific IPs is like telling a flood not to wet the floor.*

*The reality is that the market for replication is now decentralized. As Ars Technica notes, any user blocked by Claude can simply migrate to other frontier models that lack these restrictions. Even if the entire frontier AI industry agreed to a ban, developers could pivot to open-weight models that are not subject to corporate usage policies and are nearly as capable. The efficiency of replication has been unlocked; you cannot put the genie back in the bottle with a JSON file.*

### The Bottom Line

Anthropic's effort to strike a balance—allowing "original games, genre games, and games built from your own designs" while blocking clones—is a noble attempt at corporate citizenship. However, as GitHub issue reports suggest, these prompts are already interfering with legitimate game development use cases.

In the end, the "tengu_heron_brook" rule is a shield for Anthropic's legal department, not a stopgap for the industry. The AI copying boom is not a bug of the system; it is the system. As long as the incentive is to "make number go up" via rapid replication, the tide will continue to rise, regardless of how many sentences Addy Osmani adds to a system prompt.