================================================================================
        
C:\> CD ..   Return to Index
        
================================================================================
[February 1, 2025]

Brain Dump Prompting
AI, XY and the Art of the Brain Dump

* * *

[THE XY PROBLEM]
---------------
You ask someone, "how do I get to Millbank Street?" (your presumed 'Y solution') 
When what you really want to know is: "How do I get to the Pizza Post?" (your actual 'X problem') 

You get perfect directions to Millbank Street assuming the Pizza Post was on Millbank Street. But, the Pizza Post is on Putnam Avenue. 

Had you asked about your X problem you would've gotten the correct answer, but you asked about your Y solution instead. Your hangry kids are losing it and your wife's stares burn deep.

As soon as you understand the XY problem (nearly autocorrected to the 'ex-wife' problem!), you see it everywhere - business, home, tech, life. 

[THE AI CONNECTION]
-----------------
But understanding the XY problem is the unlock to understand how to use AI (let's use 'AI' here as rough placeholder for large language models). 

Most people use AI chatbots like a search engine, looking for some narrow bit of information, a business, an address, an article, a YouTube video on how to fix a leaky shower head.

But recognizing the XY problem is the unlock to AI. It is a mindset shift.

Why? Because they are designed to generate an output based on input. The more specific, tailored and detailed that input, the more 'useful' the output. 

What does this mean in practice? AI needs context, overwhelming, shock and awe context.

[ENTER: BRAIN DUMP PROMPTING]
---------------------------
Brain Dump prompting is exactly what its sounds like. It is the equivalent of speaking for 60 secs straight about the exact uncertainty you are trying to resolve. Voice-to-text dictation works best here [LINK: mac-dictation]. Explain in as much detail as you have available. 

Don't know all the angles and context? Not sure exactly how to frame the problem? That's fine, that's data right there. State exactly that.

Begin with what you do you know. This is an essential mistake I and I have seen others make. Not asking the question when we don't have all the details or think we don't have the full picture. 

Said differently. Feel like you are stuck with AI? Use AI to get unstuck. 

[BRAIN DUMP PROMPTING IN PRACTICE]
---------------------------
Let's call the old way the 'Hey Siri' method. What you would have typed into a Google search.

METHOD 1: Hey Siri
-----------------
INPUT:
> Give me a recipe for Thanksgiving turkey

OUTPUT: 
> A perfectly fine turkey recipe. 

METHOD 2: Brain Dump
------------------
INPUT:
> OK shit. I'm at my mother-in-laws house in Kansas City. I agreed to be in charge of the turkey. Regretting this deeply. It's the Wednesday before Thanksgiving and my turkey is half frozen after being in the fridge for 8 hours. All I've got laying around is half a stick of butter, the old thyme, sage, dried parsley, Morton's salt and a tin of black pepper that's probably been in here since the Carter administration. Dinner is for 12 people, 8 adults, 2 kids. We're supposed to sit down at 2. Turkey is 15 lbs. It's 830pm now. I can't have a Clark Griswold style dry turkey. I'll RUIN Thanksgiving! We need a juicy bird and gravy. I need a game plan. Let's do this. Failure is not an option.

OUTPUT: 
> A game plan. 

Is Brain Dump Prompting inefficient? The opposite: it's highly efficient. It is the first step to knowledge. You are starting with the knowledge you have today, including the known unknowns, and using AI itself to refine and structure a path to uncovering new knowledge, including the unknown unknowns. 

This is also an essential method for refining system prompts.

When in doubt: Brain Dump, refine, repeat

BTW. Greg Stove and Taylor Malmsheimer write a terrific substack that touches on AI use.
================================================================================
C:\> CD ..   Return to Index
================================================================================