Claude Code users say Anthropic’s coding agent picked up a server-delivered rule against recreating existing games. A screenshot of a user’s Claude Code config, posted on X on 30 October by @chasmmmmmmmmmmm, shows a setting named tengu_heron_brook with the text “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.” The poster says the rule arrived “around Oct 27” and that “There’s no setting to turn it off”.
The post had 451,243 views when we read it at about 13:39 UTC on 30 October. A GitHub issue opened the same morning by a developer who builds licensed board-wargame software says the rule has stopped paid client work. Anthropic had not published or confirmed the rule in the sources we checked, listed below.
Short explainer · Claude Code Game Rule Post Draws 451K Views
A vertical news Short narrated from this piece’s headline and opening line.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on the model and the harness. It teaches the general idea only and does not discuss this rule or games.
The key insight: Claude Code users report a server-delivered rule, shown in a config screenshot under the setting tengu_heron_brook, telling the agent not to build a recreation of a specific existing game; a licensed board-wargame developer says on GitHub it has stopped paid client work, and Anthropic had not confirmed the rule in the sources we checked.
The screenshot is labelled “.claude.json” and “Claude Code config”. The chasm post says the rule was “sent from Anthropic’s servers” and asks: “Is this the end of game rewrites, modding and everything in between?”
Quoting that post, @kimmonismus wrote that “Anthropic no longer allows good benchmarks like Minecraft or GTA 5 recreational games”, adding: “To be fair: probably as a consequence of the IPO, to avoid legal issues due to copyright infringement.” The IPO link is the poster’s view; neither post cites an Anthropic statement. That post had 12,398 views when we read it.
By our arithmetic, the two posts carried 463,641 views at that time, and the original had about 36 times the views of the quote post.
Views on the two X posts reporting Claude Code’s game-recreation rule, as shown by the fxtwitter API at about 13:39 UTC on 30 October 2026. Combined 463,641; about 36 times (our arithmetic). View counts change over time.
What Was Posted on X
ANTHROPIC IS CRACKING DOWN ON VIDEO GAME REWRITES ๐จ
Claude Code picked up a new rule around Oct 27, sent from Anthropicโs servers under a flag called “tengu_heron_brook”. Thereโs no setting to turn it off.
The full rule: โDo build original games, games in a genre, and games from the userโs own designs and assets. Donโt build a recreation of a specific existing game, even if it is renamed or called original.โ
Is this the end of game rewrites, modding and everything in between
@chasmmmmmmmmmmm says Claude Code picked up a new rule around Oct 27, sent from Anthropic’s servers under a flag called tengu_heron_brook, and that there is no setting to turn it off.
Another reason for open source. Unfortunately, Anthropic no longer allows good benchmarks like Minecraft or GTA 5 recreational games.
To be fair: probably as a consequence of the IPO, to avoid legal issues due to copyright infringement.
@kimmonismus calls it another reason for open source and says it is probably a consequence of the IPO, to avoid legal issues due to copyright infringement; that is the poster’s view.
Looks like Anthropic are cracking down on Claude Code videogame mashups. I’ve seen some amazing ones posted in group chats. The way this has progressed since September, it will soon be as easy as mixing together two colors of paint.
Andrew Curran (@AndrewCurran_), in his read on 30 Oct, says it looks like Anthropic is cracking down on Claude Code videogame mashups, and that he has seen some amazing ones posted in group chats.
A Licensed Developer’s Complaint
GitHub issue #106167 on Anthropic’s claude-code repository was opened at 09:25 UTC on 30 October, 2 hours 24 minutes after the chasm post by our arithmetic. Its author writes: “Around October 27, Claude Code began receiving a server-side instruction that causes sessions to refuse game-related development work.”
The author says they run Conflict Simulations LLC, which “builds digital modules of board wargames under paid commission from the clients and publishers who own those games”, and argues that “A blanket category rule against “games” cannot tell the difference between piracy and a licensed professional implementation”.
In a later comment the author says Anthropic’s support bot, on two accounts, stated the policy “applies to all customers without exception”, and that a Safeguards appeal has been filed. The issue asks Anthropic to “Document server-delivered instruction changes of this scope in the Claude Code changelog”. When we read it, the issue was open with no comment from an Anthropic account.
The Flag Has Carried Other Text
The flag name is not new. GitHub issue #80988, opened on 24 July, reports that Claude Code 2.1.219 injected a system-prompt section registered as heron_brook with different text, “Do not call the AgentTool unless the user requested it”, and says “there is no documented way to turn it off”. A search of the repository on 30 October for heron_brook returned 11 issues.
The IP Backdrop
Anthropic’s usage policy already bars using its products to “Infringe, misappropriate, or violate the intellectual property rights of a third party”. Its S-1, as our earlier coverage reported, shows $1.2 billion of its $1.5 billion Bartz copyright settlement recorded as a current liability at 30 June. As we also reported, the filing says Anthropic may indemnify “without a cap on our liability” against certain claims that its models or their outputs infringe intellectual property.
As we read the reports, the issue the developer raises is less the rule’s aim than its reach: the GitHub author argues a category rule against games cannot tell piracy from licensed work.
As we read the GitHub record, the same setting name carried different text in July, when issue #80988 reported a heron_brook section about the AgentTool with no documented opt-out.
As we read it alongside our S-1 coverage, the rule sits next to an IP backdrop Anthropic has already disclosed: $1.2 billion of the Bartz settlement still recorded as owed at 30 June.
Rule text in a user’s Claude Code config screenshot, posted by @chasmmmmmmmmmmm on X, 30 October 2026
“Don’t build a recreation of a specific existing game, even if it is renamed or called original.”
Three Implications
FOR DEVELOPERS The GitHub reporter asks for a verification or exemption path for licensed work, comparable to the existing program for vetted security researchers.
FOR ANTHROPIC The issue asks it to document server-delivered instruction changes of this scope in the Claude Code changelog; the changelog we read had no entry about games.
FOR THE OPEN-WEIGHT DEBATE @kimmonismus calls the rule another reason for open source, a view the post states without citing Anthropic.
The framework says “The surface, not the model, governs autonomy.”
As we read it, a rule pushed to the harness from the server is an example of the surface setting the limits. The framework does not discuss this rule or games.
What Is Not Established
Anthropic had not confirmed the rule, its scope or its timing in the sources we checked: the Claude Code changelog (newest entry 2.1.314, with no entry about games), its usage policy page and GitHub issue #106167. We did not contact Anthropic. The rule text comes from a user’s screenshot, which we could not verify independently. The reported start date of around 27 October is the users’ account, and the IPO motive is @kimmonismus’s view. How many users or projects are affected is not stated in any source we read.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Claude Code users report a server-delivered rule, shown in a config screenshot, telling the agent not to build a recreation of a specific existing game. A licensed board-wargame developer says on GitHub that it has stopped paid client work, and as of our reading Anthropic had not published or confirmed the rule.
A note on sourcing. The rule text comes from a user’s screenshot posted on X; we could not verify it independently and Anthropic had not confirmed it in the sources we checked. GitHub facts are the reporters’ own accounts. S-1, Cursor, Jules and Wikimedia facts come from our earlier coverage. View counts are as shown when we read them. Figures labelled our arithmetic are our own calculations.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
