A prompting technique where I externalize thinking without directly addressing the LLM. The agent responds to the thought stream rather than engaging in conversation.

So instead of “talking to” an AI, I’m thinking out loud and something picks up on what’s actionable. The agent becomes environment rather than entity (like how a good IDE responds to what I’m doing without explicit commands).

Why This Works

Standard prompting has its place, but it’s a conversation. With reactive prompting, I’m talking to myself. The agent overhears and responds to what’s actionable. The thoughts are the source of truth, not a back-and-forth.

What Triggers Response

Not everything in the stream needs a response. The triggers I’ve noticed:

• Expressed need or uncertainty
• Ambiguity that blocks progress
• Errors or contradictions worth flagging
• Tasks implied but not stated

Pure reflection can flow past without interruption.

The Keywords Hint

When the thought stream needs specific context or tools, I add a keywords: line at the end:

Discussion about this needed x-devonthink-item://...

keywords: devonthink mcp, mcporter

The keywords act as hints for tooling or skills. It’s metadata for the stream, not a command.

Distill is a product built on this model. AI agents watch threads, spot patterns, and act without being prompted.