The secret life of unnamed streets in your AI stack
A system to help you stop losing your best ideas in a mess of random chats by building a City Map for your AI through strategic interrogation.
Welcome to todayâs edition of ROBOTS ATE MY HOMEWORK. Today, weâre diagnosing why your smartest AI workflows keep vanishing, then building them street names so they stick around long enough to compound.
A few months ago, I picked up Deirdre Maskâs The Address Book. She outlines that homes without addresses effectively disappear. Ambulances canât find you, job offers never arrive, and you vanish from the parts of the world that create opportunity.
Around the same time I read this book, Paris did the exact opposite of erasing someone. It named a viewpoint in Montmartre the âLa BohĂšmeâ Belvedere in honor of Charles Aznavour, one of my all-time favorite musicians.â

These two things (a missing address in the middle of nowhere and a named belvedere in Paris) are the extremes of the same idea: naming and addressing determine value.
An address allows a location to accumulate history and compound its worth over time. Unaddressed locations simply exist in the void.
Your prompts work the same way.
An organized Prompt City functions as a transferable business asset. You can hand it to a junior employee and expect senior-level output immediately. Disorganization, conversely, creates a liability.
Right now, most of us suffer from Prompt Rot. Our smart strategies quietly decay because we fail to give them a permanent address.
Our attempts to build an AI stack have resulted in accidental urban sprawl. Lacking zoning laws or a transit system, this sprawl costs us 5â10 hours a week in âre-creation time.â
Today, you become the Urban Planner through a series of increasingly uncomfortable questions that force you to see what youâve been avoiding.
Hereâs what we're building:
A diagnostic system that reveals your real priorities (not your stated ones)
The Ghost District Protocol: finding the gap between what you say matters and where you invest your time
A naming convention that turns chaos into reusable intellectual property
One 5-minute audit that delivers immediate value
To keep this doable instead of overwhelming:
Start with one domain (e.g., ânewsletter writingâ), not your entire AI history.
Youâll run just 4 prompts, each more revealing than the last
No exports, no spreadsheets, no admin work. Just plain old thinking
Letâs give your workflows street names.
đ± Step 1: Reveal the âDark Cityâ
Look, your first instinct is probably to export your entire AI history because data is king, right? More data equals more insights.
Except thatâs a trap. Youâll end up with a massive file that takes two hours to sort through and youâll never do anything with it.
Instead, weâre doing something simpler. Weâre going to look at your most recent behavior, the stuff youâve been doing in the last couple weeks without thinking about it.
Open your AI history sidebar. Look at the last 10 chats (or more if youâve got some more time on your hands) you spent a lot of time on.
For each one, write down a super short description. Like 3-5 words max.
Mine are super random and look like:
âany actionable stuffâŠâ
âSYSTEM ROLE & IDENTITYâŠâ
âi need help figuring outâŠâ
âi need you to be a bit objâŠâ
âTHE ROBOTS METHODâŠ.â
And yeah, I still use âI need you to be a bit objectiveâ prompts đ
Donât clean this up. Donât skip the embarrassing ones where you asked AI to write your grocery list or whatever. This is your raw data, and raw data doesnât lie.
Now take that list and run this prompt:
Iâm going to show you the last 10 things I used AI for. Your job is to tell me the uncomfortable truth about my priorities.
Here are my last 10 AI uses:
[paste your list]
Analyze this and tell me:
1. What am I over-indexing on? (Where am I spending way too much energy on stuff that probably doesnât move the needle?)
2. Whatâs the real pattern here? (Are all these different requests just me doing the same thing over and over?)
3. What strategic work is completely missing from this list?
4. If this is how Iâm using AI, what does it say about what Iâm avoiding?
Be direct. I need the truth, not a pat on the back.This is going to show you the gap between what you say your priorities are and what your behavior reveals they are.
đ± Step 2: The Ghost Districts
Okay, so now youâve got the AIâs analysis from Step 1. And if youâre anything like me, thereâs at least one thing on that list that made you go âouch.â
Look specifically at question #3 - the strategic work thatâs completely missing.
Thatâs your Ghost District.
Ghost Districts exist because youâve been actively avoiding building there. Maybe itâs scary. Maybe you donât know how to do it.
Weâre going to diagnose why by asking better questions.
Run this follow-up prompt:
You just told me that [INSERT THE MISSING STRATEGIC AREA FROM STEP 1] is completely absent from my AI usage.
I need you to help me understand why. Act as a strategic psychologist and diagnose this gap.
Ask me 5 questions that will reveal whether this is:
- A skill gap (I genuinely donât know how to do this)
- A confidence gap (I know how, but Iâm terrified of being bad at it)
- A clarity gap (I donât know what âgoodâ looks like here)
- A priority gap (I say it matters, but I donât really believe it does)
- Or a systems gap (I want to do it, but I have zero infrastructure for it)
For each question, explain what my answer would tell you about the real problem.Important bit: You need to actually answer these questions. Out loud if you can, or in writing if youâre not ready to hear yourself say it yet.
You probably have never thought of doing this before, did you? Itâs a trivial task that most people wouldnât spend time on⊠But this is also where the money is and where your actual business growth might be hiding.
So donât skip it.
đ± Step 3: Build your prompt pattern library
Alright, enough pain.
Go back to your âLast 10 Clicksâ list from Step 1. This time, look for the 3-5 chats that led to something you shipped. Things you actually sent, published or implemented.
These are your Main Arteries.
The routes through your AI city that lead somewhere.
Thing is, most people think they have dozens of different workflows. But when you look at what works, you usually have like 3-5 core patterns that youâre just applying to different situations.
Letâs extract those patterns. Run this prompt:
Iâm going to show you the 3-5 AI conversations that led to completed, shipped work in the past month.
[Paste a brief description of each oneâjust 2-3 sentences about what you asked AI to do and what you ended up shipping]
Your job is to find the reusable pattern hiding underneath.
For each conversation, tell me:
1. What job was I hiring AI to do? (Use that Jobs-to-be-Done framework)
2. What specific type of thinking did AI help me do? (Generate options? Critique my draft? Force me to clarify my fuzzy thinking? Synthesize a bunch of inputs?)
3. If I turned this into a template, what parts change each time [MARK THESE AS VARIABLES] and what parts stay the same?
4. How did I know the output was good enough to ship?
Show me the architecture underneath my chaos.What youâre probably going to discover is that your â50 different promptsâ are just like 3-5 core jobs wearing different costumes. Once you see that, everything gets easier.
Because instead of reinventing the wheel every time, you can just reuse the pattern with different variables.
đ± Step 4: The Metro Protocol
Hereâs where we give your patterns permanent addresses. Because right now, even if you know you have a good prompt somewhere, finding it is like trying to remember where you parked at the airport.
The Metro Protocol is dead simple: [City] / [District] / [Street] / [Version]
Think of it like:
City = Your company or brand name (so if you have multiple projects, they donât get mixed up)
District = The big strategic bucket (Research, Drafting, Editing, Distribution, Strategy, Operations)
Street = The specific type of workflow
Version = Which iteration youâre on
So instead of âthat prompt I used last month for competitor research,â youâd say: RAMH / Research / Competitor Intel / v3
See the difference? One requires you to dig through your memory. The other is an actual address you can just... go to.
Take your 3-5 patterns from Step 3 and run this prompt:
Iâve identified my core AI patterns. Now I need to give them proper addresses using the Metro Protocol format: [City] / [District] / [Street] / [Version].
Here are my patterns:
[Paste the output from Step 3]
For each one:
1. Give it a Metro Protocol address that makes sense
2. Rewrite the prompt as a template where the stuff that changes is marked with [BRACKETS]
3. Write a one-sentence âWhen to Useâ guide so I know when to pull this address
4. Show me an example of the template with real variables filled in
My city name is: [YOUR BRAND/COMPANY NAME]
Make the street names memorable and specific. I donât want generic labels like âContent Creation.â I want to know exactly what this does.What youâll get back is your first official set of named, reusable workflows. You donât need a fancy system for this. Just save them in a Google Doc or wherever you keep important stuff.
đ± Step 5: The 5-minute audit
Look, I know some of you are reading this thinking âthis is great but I donât have time for all these steps right now.â
Fair. So hereâs your shortcut.
If you do nothing else today, do this one thing:
Copy the first 2-3 sentences from your last 5 AI conversations. Doesnât matter which domain, doesnât matter if theyâre good or bad. Just grab them.
Then run this prompt:
Iâm going to give you my last 5 AI requests. Tell me:
1. What Iâm over-indexing on
2. What strategic area is completely missing
3. The ONE workflow I should name and standardize first
Here are my last 5 requests:
[paste them]
Be specific. Give me the exact address I should create today.This takes 5 minutes and it tells you where to start. You donât need to build the whole city today. Just build one street and see how it feels.
I bet youâll get addicted â€ïž
đ± Step 6: The Ghost District builder
Remember your Ghost District from Step 2? That strategic area thatâs been conspicuously absent from your AI usage?
Yeah, weâre not letting that slide anymore.
Hereâs your protocol for the next 7 days:
Every time you open ChatGPT or Claude, before you type anything, ask yourself: âIs this request in my Ghost District?â
If yes: Create an official Metro Protocol address for it FIRST. Then run the prompt.
If no: Use an existing address if you have one, or just acknowledge itâs a one-off experiment.
Whatâs going to happen is youâll start catching yourself. Youâll notice patterns youâve been blind to. Youâll realize that half your âurgent, unique, one-timeâ requests are just variations of the same 3-4 patterns.
Youâll see in real time where youâre lying to yourself about your priorities.
Because the truth is your Ghost District is missing because youâve been choosing to avoid it.
And now you canât hide from that anymore.
Congrats, your AI prompt city just got an upgrade
Deirdre Mask showed that addresses arenât neutral. They decide who gets counted, who gets services, who gets found, and who holds power.
Your AI address system does the same thing.
It decides which workflows get counted as official strategy, which ones get the infrastructure to run smoothly and which ideas get found when you need them.
Youâve just mastered âstrategic visibility.â
When every prompt is addressed by district and purpose, you can see at a glance where youâre over-investing and under-investing relative to your stated goals. If you say youâre focused on thought leadership but your Strategy District has two addresses while your Operations District has twenty, your AI city is betraying your real priorities.â
You are actively building a proprietary operating system.
And that saves you TIME, FOCUS and HEADACHES. Youâre welcome!
So, Mayor, which Ghost District are you going to build first?
To naming your forgotten streets,
Chief đ€ at ROBOTS ATE MY HOMEWORK
đ Feel like you just discovered a new district? These five archived addresses might be the next blocks you want to explore:














Prompt City as real IP where once the streets are named, you can onboard someone into nearâsenior output in days instead of years.
I love how you don't just give instructions on how to use AI but practical real examples that all of us that interface with AI daily do. This truly helps to build confidence.
Thank you for this article.