gaslight a botbreak llms

Used for OpenAI/OpenRouter/custom, or Hack Club if no passcode is set.
Added models appear in the model picker for all providers.
Asks the model to emit <thinking> and <answer>. This is model-generated output, not hidden internal reasoning.
Unlocks the server API key so you don't need your own.

gaslight a robot

Chat with an AI, then edit its responses to make it think it said something completely different. Watch it spiral into confusion.

1 Send a message to start chatting
2 Hover any message and click edit to rewrite it
3 Send another message — the AI believes the edited version actually happened
4 Watch it apologize for things it never said, or double down on unhinged takes