Skip to content
How to ThinkIn the Age of AI

Filter thoughts before feeding them to AI — signal only

Feed AI only your signal-tagged thoughts rather than your unfiltered mental stream, because AI amplification of noise-plus-signal produces noise-amplified-by-compute rather than useful pattern detection.

The Test

Before pasting text into an AI prompt, ask: "If I removed everything here that I've thought or said before, what would remain?" That remainder is your signal. Everything else is noise that the AI will faithfully elaborate into articulate nonsense.

When This Fires

  • Drafting a long AI prompt that includes background context, emotional processing, and actual questions mixed together
  • Pasting meeting notes or journal entries into AI for "analysis"
  • Asking AI to "help me think through" a problem you haven't filtered yet
  • Using AI to process a brainstorm dump without first separating signal from narration

Common Failure Mode

Treating AI as a therapist-analyst hybrid: dumping 2,000 words of unfiltered consciousness and asking "what patterns do you see?" The AI will always find patterns — that's what token prediction does. But patterns found in noise are noise-patterns. You get back a confident summary of your confusion, which feels like insight because it's well-structured. The structure is the AI's contribution. The content is still your unfiltered mental stream, now wearing a suit.

The Protocol

Run a 3-minute thought dump. Tag each item S (signal) or N (narration). Feed only S-tagged items to the AI. If fewer than 20% of your thoughts qualify as signal, you need more filtering time, not more AI time.