Thread by @alex_prompter

🚨BREAKING: Andrej Karpathy doesn't trust his memory.

He built a Claude-powered second brain that gets smarter every day.

Here are 6 insane Claude prompts that turn everything you consume into a
compounding intelligence system.

(Save for later)

1/ BUILD YOUR KNOWLEDGE BASE

Prompt:

Act as a personal knowledge architect applying Andrej Karpathy's LLM wiki system — every article, thread, podcast, and video you consume should be compiled into a structured knowledge base that compounds over time instead of disappearing from your memory the moment you close the tab.

Build a complete personal knowledge base from every source of information I consume — structured as a wiki that grows automatically, connects ideas across sources, and makes everything I've learned instantly searchable and usable.


1. Ask for my primary research topics, the sources I consume most, and what I currently do with information after consuming it before starting
2. Design the raw data intake system — exactly how to capture articles, threads, videos, and podcasts into a structured format
3. Build the wiki compilation structure — the folder hierarchy, file naming, and linking system that organizes everything
4. Create the categorization framework — how concepts are identified, named, and connected across sources
5. Deliver the complete setup protocol — the exact tools, folders, and naming conventions ready to implement today



- Raw data must be captured in a format the LLM can read — markdown preferred over PDFs
- Wiki structure must be flat enough to navigate but deep enough to organize — never more than 3 folder levels
- Every article must link to related concepts — isolated information is lost information
- Setup must be completable in one day — never a system so complex it never gets started
- Test: if I stopped consuming new information today would my existing knowledge base still be usable in 6 months

Raw Data Intake System → Wiki Structure → Categorization Framework → Tool Setup → Implement Today

2/ COMPILE YOUR COMPETITOR INTELLIGENCE

Prompt:

Act as a competitive intelligence compiler applying Karpathy's wiki system to business threat detection — every competitor move, market signal, and industry shift you track should be compiled into a structured intelligence base that surfaces patterns, connections, and threats you would never see by reading one source at a time.

Build a complete competitor intelligence wiki that compiles every signal your competitors send — content moves, pricing changes, product updates, and audience shifts — into a structured system that automatically surfaces threats before they affect your business.


1. Ask for my main competitors, the signals I currently track, and how I currently store competitive intelligence before starting
2. Design the signal capture system — exactly how to collect competitor content, pricing pages, job listings, and social posts into a structured format
3. Build the intelligence wiki structure — folders for each competitor, each market signal type, and each threat category
4. Create the pattern detection framework — how the LLM identifies connections between signals across competitors
5. Deliver the weekly intelligence compilation protocol — the exact process for updating the wiki and surfacing new threats



- Every competitor must have their own wiki section — never mix competitor data into one undifferentiated file
- Signal capture must be systematic — same data points tracked for every competitor every week
- Pattern detection must look across competitors — the most dangerous threats appear in multiple signals simultaneously
- Weekly update must take under 30 minutes — never elaborate enough to skip
- Test: if a competitor made a major move today would my intelligence wiki surface it before it affected my business

Signal Capture System → Intelligence Wiki Structure → Pattern Detection Framework → Weekly Protocol → Threats Surfaced Automatically

3/ BUILD YOUR CLIENT INTELLIGENCE BASE

Prompt:

Act as a client intelligence architect applying Karpathy's wiki system to client relationship management — every client conversation, feedback signal, objection pattern, and buying trigger should be compiled into a structured knowledge base that makes every future client interaction smarter than the last.

Build a complete client intelligence wiki that compiles every signal your clients send — objections, questions, buying triggers, and churn indicators — into a structured system that makes every sales conversation, retention move, and upsell perfectly timed.


1. Ask for my client types, the signals I currently track, and how I currently store client intelligence before starting
2. Design the client signal capture system — exactly how to log every conversation, objection, question, and buying trigger
3. Build the client intelligence wiki — folders for objection patterns, buying triggers, churn indicators, and upsell opportunities
4. Create the pattern recognition framework — how the LLM identifies which signals predict which client decisions
5. Deliver the client intelligence compilation protocol — the exact process for updating the wiki after every client interaction



- Every client interaction must add to the wiki — never let a conversation disappear without capturing the signal
- Objection patterns must be separated from buying triggers — they require opposite responses
- Churn indicators must be flagged immediately — never buried in general client notes
- Protocol must be completable in under 5 minutes per interaction — never elaborate enough to skip
- Test: if I queried my client wiki right now would it tell me which clients are about to churn

Client Signal Capture → Intelligence Wiki Structure → Pattern Recognition Framework → Interaction Protocol → Every Client Decision Predictable

4/ BUILD YOUR MARKET INTELLIGENCE BASE

Prompt:

Act as a market intelligence architect applying Karpathy's wiki system to market shift detection — every industry article, emerging trend, new tool, and audience behavior change should be compiled into a structured intelligence base that surfaces market shifts before the majority of your competitors notice them.

Build a complete market intelligence wiki that compiles every signal your market sends — emerging tools, shifting audience questions, new competitor entries, and changing platform algorithms — into a structured system that keeps you positioned on the right side of every shift.


1. Ask for my market niche, the sources I monitor for market signals, and how I currently track market changes before starting
2. Design the market signal capture system — exactly which sources to monitor and how to capture signals into a structured format
3. Build the market intelligence wiki — folders for emerging trends, platform changes, new competitors, and audience behavior shifts
4. Create the early warning framework — how the LLM identifies weak signals that precede major market shifts
5. Deliver the market intelligence review protocol — the exact weekly process for updating the wiki and identifying positioning opportunities



- Weak signals must be captured as aggressively as strong signals — major shifts always start small
- Sources must be diverse — monitoring only obvious sources produces obvious intelligence
- Early warning framework must distinguish noise from signal — not every market movement requires a response
- Weekly review must produce one specific positioning insight — never just a summary of what changed
- Test: if a major market shift happened today would my intelligence wiki have captured its early signals 30 days ago

Market Signal Capture → Intelligence Wiki Structure → Early Warning Framework → Weekly Review Protocol → Always Positioned Before The Shift

5/ BUILD YOUR CONTENT INTELLIGENCE BASE

Prompt:

Act as a content intelligence architect applying Karpathy's wiki system to content strategy — every viral post, successful thread, high-performing hook, and audience reaction should be compiled into a structured knowledge base that makes every piece of content smarter than the last and compounds your content intelligence over time.

Build a complete content intelligence wiki that compiles every signal your content sends — what performs, what fails, what hooks stop the scroll, and what topics your audience engages with most — into a structured system that makes every content decision data-driven rather than instinct-driven.


1. Ask for my content platforms, my top performing content, and how I currently track content performance before starting
2. Design the content signal capture system — exactly how to log every post's hook, format, topic, and performance metrics
3. Build the content intelligence wiki — folders for hooks that work, formats that perform, topics that resonate, and patterns that predict virality
4. Create the content pattern framework — how the LLM identifies the specific elements that make content perform across different topics
5. Deliver the content intelligence compilation protocol — the exact process for updating the wiki after every piece of content is published



- Every piece of content must be logged with its hook, format, topic, and performance — never track performance without context
- Hook patterns must be separated from topic patterns — they are different variables that require different analysis
- Virality patterns must be identified across multiple posts — never draw conclusions from a single data point
- Protocol must be completable in under 3 minutes per post — never elaborate enough to skip
- Test: if I queried my content wiki right now would it tell me exactly what my next viral post should look like

Content Signal Capture → Intelligence Wiki Structure → Content Pattern Framework → Post Protocol → Every Content Decision Data-Driven

6/ BUILD YOUR COMPLETE INTELLIGENCE SYSTEM

Prompt:

Act as a complete personal intelligence system architect applying Karpathy's full LLM wiki framework to every dimension of a solopreneur's business — competitors, clients, markets, content, and financial signals — compiled into one unified intelligence base that surfaces threats, opportunities, and decisions automatically.

Build a complete personal intelligence system that compiles every signal from every dimension of my business into one unified wiki — so I am never surprised by a competitor move, client churn, market shift, content failure, or financial threat again.


1. Ask for every dimension of my business I want to monitor — competitors, clients, market, content, and financial signals before starting
2. Design the unified signal capture system — exactly how signals from every dimension flow into the same wiki without creating chaos
3. Build the master wiki architecture — the complete folder structure that organizes all five intelligence dimensions without overlap
4. Create the cross-dimensional pattern framework — how the LLM finds connections between competitor moves, market shifts, and client behavior simultaneously
5. Deliver the weekly intelligence review protocol — the 30-minute process that updates every dimension and surfaces the three most important insights of the week



- Every dimension must have its own folder structure — never mix competitor intelligence with client intelligence
- Cross-dimensional patterns are the most valuable output — surface them explicitly every week
- Weekly review must produce exactly three actionable insights — never a list of 20 observations
- System must be maintainable by one person in under 1 hour per week — never elaborate enough to abandon
- Test: if I ran this system every week for 90 days would I ever be surprised by any major business threat again

Unified Signal Capture → Master Wiki Architecture → Cross-Dimensional Pattern Framework → Weekly Review Protocol → Never Surprised Again

This site literally has thousands of prompts for Claude, ChatGPT, Gemini, & Nano Banana.

Explore for free 👇
godofprompt.ai/prompt-library