Remove response: Pruning chats helps outputs
Skill spotlight – Madgwick
The Remove Response feature in Madgwick helps you maintain cleaner, more focused conversations by deleting unwanted AI responses while preserving your original input for editing.
Why use it?
Context Degradation Syndrome (CDS) is a term for the tendency of LLM outputs to degrade over long chats. Research shows even high context window AI models begin losing effectiveness between 16,000-64,000 tokens, leading to less accurate responses. Remove Response allows you to prune your chat to ensure all text being processed is useful and efficient.
How remove response works
Hover below the AI response to show the Remove button
Click the Remove button and then select Confirm to delete the AI's answer
Your original input text remains* in the message field, ready for editing
*Tip: if removing more than one response at a time, only your most recent input will be retained, so copy your previous chats if you plan to amend and reuse them.
When to use
Quick fix – when you can see what part of your input has caused an off-topic response, or you’ve accidentally hit return before finishing your input.
Strategic conversation cleanup - for off topic ‘side-quests’, or rabbit holes that prove unhelpful or distracting.
Summary – in a long chat where the detail is not crucial use a summary prompt (eg: "Review our conversation starting from '[insert specific text]'. Create a succinct summary filtering out failed experiments and tangents"), remove the lengthy back-and-forth,and cut and paste in the summary in its place.