For anthropic models it is encouraged to prefill claude's responses for improved behavior.

(https://docs.anthropic.com/claude/docs/prefill-claudes-response)

It would be great if we could have the ability to use this. We would need to be able to create dummy ai chat messages (which also is something that is generally helpful), as well as be able to choose to "continue" rather than regenerate or add a new chat message. This feature will also be usable by other models such as many of those found on openrouter.

Ideally, it would also be great to have this implementation also be compatible with AI commands, though that would require the ability to lock which model you are using (for instance, you wouldn't be able to, as far as I am aware, use a command with a prefilled response with chatgpt.)