Claude Code gets a server-sent rule against cloning games
chasm found the flag, Addy Osmani confirmed Anthropic added two sentences, Peter Steinberger used it to argue for your own harness.
Claude Code picked up a new rule around Oct 27 that was sent from Anthropic's servers under a flag called "tengu_heron_brook", according to chasm, who posted the text of it. There is no setting to turn it off.
The rule, as chasm quoted it, tells the model to build original games, games in a genre, and games from the user's own designs and assets, and not to build a recreation of a specific existing game, even if it is renamed or called original.
Addy Osmani confirmed the change in a reply. He said two sentences were added to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game, because using Claude to infringe IP is against Anthropic's usage policy. Original games, genre games, and games built from your own designs and assets are meant to keep working, he wrote, and Anthropic will keep iterating to get the balance right. If Claude turns something down that is yours, he asked people to send it his way.
Original games, genre games, and games built from your own designs and assets are meant to keep working.
The part that is not about copyright
Peter Steinberger's read skipped the IP question entirely. He called it another reason to use your own harness rather than one where companies change what you can and cannot do at will.
That is the bit worth sitting with. The behaviour of the coding agent on your machine changed without a release, without a changelog entry you went looking for, and without a toggle. Somebody noticed a flag name in the traffic. chasm asked whether this is the end of game rewrites and modding, which nobody in the thread answered, and the honest status is that the only published description of the new boundary is the two sentences themselves plus Osmani's assurance about what should still work.
For working engineers, the practical takeaway is narrow. If your agent starts refusing a task it used to do, the explanation may not be the model and may not be your prompt. It may be an instruction that arrived from a server since the last time you ran it, which is a different debugging problem than the one most people are set up for.
Another reason why you wanna use your own harness and not one where companies change what you can or can鈥檛 do at will.
Have the good opinion, not the loud one
One email a day. Under three minutes. What working engineers found out about coding with AI.