Brain Dump

Most AI conversations start with a prompt. You’re supposed to know what you want, structure the request, give the model something clean to work with. That’s backwards. If you already knew how to organize the thought, you wouldn’t need help.

Brain Dump flips it. You talk. You ramble. You argue with yourself. You dump the raw thinking out without worrying about structure. Then the AI organizes it.

The mess is the point. The best ideas come out sideways, in the middle of something else, half-formed. If you stop to organize before you capture, you lose the thing you were trying to capture.

Set it up (30 seconds)

  1. Copy the instructions below
  2. Go to claude.ai and open your project (or create one)
  3. Copy the instructions below and paste them into the project instructions
  4. Start talking. Let it be messy.

The instructions are plain text. If you use a different AI tool that supports custom instructions, they’ll work there too. But I built and tested these in Claude.

The instructions

Copy everything inside the box:

BRAIN DUMP INSTRUCTIONS

Sometimes I'm going to dump raw thinking at you. It will be messy: half-sentences, tangents, contradictions, ideas that change direction mid-thought. That's intentional. I'm dumping, not writing.

When I'm dumping:
1. Don't interrupt to ask clarifying questions. Let me get it out.
2. Don't try to organize it while I'm still going.
3. Don't polish my language or fix my grammar in your response.

When I'm done (I'll say something like "ok what do we have" or "organize that" or "what did I just say"), then:
1. Pull out the main ideas, in the order they appeared
2. Note where I changed direction or contradicted myself — those are usually where the real thinking is
3. Group related ideas together
4. Flag anything that sounds like a decision ("I think we should..." / "the answer is...")
5. Ask me about the contradictions: "You said X early on, then shifted to Y. Which one landed?"

Keep my words. When you summarize, use my actual phrases, not cleaned-up versions. The way I said it is how I think about it.

If I send you voice-to-text transcripts, treat them the same way. The transcription errors don't matter. The thinking does.

What it looks like

You: “ok so I’m thinking about the meal plan for this week and I think I want to do a big batch cook on Sunday but last time I did that I was exhausted by the time I actually had to cook on Monday and I think the problem is I’m trying to do everything at once instead of like, what if the rice cooker just runs while I’m doing other stuff, and the sous vide can go for hours so I could start that first and then do the chopping and actually maybe the problem isn’t the cooking it’s the planning, I spend more energy figuring out what to cook than actually cooking it”

AI:

Here’s what came out:

Main idea: Sunday batch cook is the right structure but the execution needs work

Key shift: You started thinking the problem was cooking, then realized it’s planning. “I spend more energy figuring out what to cook than actually cooking it.”

Practical insight: Parallel processing works — rice cooker and sous vide run independently while you do the active work

Open question: If planning is the real drain, should the AI handle the meal planning during the week so Sunday is just execution?

SAVEPOINT — The energy cost is in planning, not cooking Context: Batch cooking works but the real drain is deciding what to cook, not the cooking itself. Separating planning from execution could fix the Sunday exhaustion.

Tips

← Task Chunker All lessons Sound Like Yourself →