Spencer Baggins Profile picture
Apr 25 14 tweets 11 min read Read on X
AI can now run competitive intelligence like a CIA analyst (for free).

Here are 12 Perplexity prompts that build the kind of threat assessments governments pay millions for.

(Save for later) Image
1/ The Signals Collection Brief

CIA analysts don't start with conclusions.

They start by collecting every available signal from every available source before forming a single opinion.

Most business intelligence fails here. People collect the signals that confirm what they already believe and call it research.

"Run a full signals collection brief on [competitor / market / geopolitical situation]. Pull from every available public source simultaneously: news from the last 90 days, regulatory filings, job postings, executive statements, patent applications, partnership announcements, customer reviews, social media activity, and any academic or industry research mentioning them. Do not analyze yet. Present everything organized by source type. Flag anything that appeared in only one source single-source intelligence is always treated with higher skepticism in CIA methodology."

The quality of your analysis is entirely determined by the quality of your collection.

This prompt collects everything before touching a single conclusion.Image
2/ The OSINT Deep Dive

OSINT Open Source Intelligence is the CIA's most underrated discipline.

80% of what intelligence agencies know about foreign governments comes from public sources that anyone could access.

The advantage isn't access. It's knowing what to look for and where.

"Run a full OSINT deep dive on [target company, person, or organization]. Go beyond the obvious sources. Pull: corporate registry filings and ownership structure changes, court records and litigation history, regulatory actions and compliance violations, property and asset records, archived versions of their website showing how their messaging has evolved, conference appearances and what topics they've spoken on publicly, and any technical footprint visible through job postings or open source code repositories. Organize findings by confidence level: confirmed, probable, and possible."

The CIA calls unverified intelligence 'raw.' This prompt turns raw into confirmed.Image
3/ The Threat Assessment Matrix

Every CIA country brief starts with a formal threat assessment.

Not a list of risks. A structured matrix that evaluates each threat on two dimensions simultaneously: probability of occurrence and severity of impact if it does occur.

Most business risk analysis is a list. Lists treat all risks as equal.

The matrix forces prioritization.

"Build a formal threat assessment matrix for [your company / market position / specific initiative]. Identify every credible threat across these categories: competitive threats from existing players, competitive threats from new entrants, technology disruption threats, regulatory threats, macroeconomic threats, and supply chain or partner dependency threats. For each threat: rate probability from 1 to 5 and impact from 1 to 5. Plot the matrix. Identify the quadrant with high probability and high impact these are your priority intelligence requirements. What signals would tell you each high-priority threat is accelerating?"

The threat you don't prioritize correctly is the one that catches you unprepared.Image
4/ The Competitor HUMINT Simulator

HUMINT Human Intelligence is what the CIA collects from actual people.

In business, your HUMINT sources are hiding in plain sight.

Former employees. Customer reviews. Conference talks. Podcast interviews. LinkedIn posts from their team.

Every one of these is a human source broadcasting intelligence about your competitor's internal reality.

"Simulate a HUMINT collection operation on [competitor]. Pull every available signal from human sources: Glassdoor and Blind reviews from current and former employees focusing on what they reveal about internal culture, strategic priorities, and operational problems. LinkedIn activity from their leadership team showing what they're thinking about publicly. Podcast and conference appearances where executives spoke candidly. Any interviews where they discussed strategy, challenges, or future plans. Synthesize what the human sources reveal about the internal reality of this organization that their official communications never show."

Official communications tell you the story they want told.

Human sources tell you what's actually happening inside.Image
5/ The Pattern of Life Analysis

CIA analysts build what they call a Pattern of Life a map of how a target behaves consistently over time.

Because behavior patterns reveal intentions more reliably than statements do.

A country that consistently moves troops to a border is planning something regardless of what their foreign minister says.

A competitor that consistently hires in a specific function is building something regardless of what their press releases announce.

"Build a Pattern of Life analysis for [competitor] over the last 18 months. Track: their hiring patterns by function and seniority over time, their pricing behavior and any changes, their product release cadence and what categories they've released in, their geographic expansion sequence, their partnership and acquisition activity, their marketing spend signals visible through ad libraries and content volume, and their executive communication frequency and topics. What behavioral patterns emerge that are consistent regardless of what they say publicly? What does the pattern predict about their next move?"

Patterns don't lie. Press releases do.Image
6/ The Red Cell Analysis

The CIA's Red Cell is a small unit of analysts whose entire job is to think like the adversary.

Not to understand the adversary. To actually become them to reason from their objectives, their constraints, and their worldview and produce the analysis they would produce.

The output is always uncomfortable. That's the point.

"Run a Red Cell analysis on [competitive situation]. You are now the senior intelligence analyst at [competitor]. You have full knowledge of your own organization's strengths, strategic priorities, and resources. You are analyzing [your company] as your primary competitive threat. From their perspective: what do they see as your weakest points right now? What move against you would produce the highest impact at the lowest cost to them? What are they probably planning that you haven't anticipated? What would they do if they knew everything about your current strategy? Write this analysis in first person as their analyst, not as an outside observer."

The most valuable intelligence is always the analysis your opponent would write about you.

This prompt forces you to read it.Image
7/ The Indications and Warning System

The CIA's Indications and Warning system is designed for one purpose.

To identify the earliest possible signals that a significant event is developing before it becomes visible to everyone.

By the time an event is obvious, it's too late to prepare for it optimally.

The competitive advantage lives in the warning period.

"Build an Indications and Warning system for [your industry / competitive situation]. First: identify the 5 most significant events that could materially change your competitive position in the next 12 months a competitor launching in your core market, a regulatory change, a technology shift, a major customer defection, a new entrant with significant backing. For each scenario: what are the earliest possible indicators that this is developing the signals that would appear 3 to 6 months before the event becomes public knowledge? Build a monitoring protocol that tells me exactly what to watch weekly to get maximum warning time before each scenario materializes."

The CIA's job is to prevent surprise.

This prompt builds the system that prevents yours.Image
8/ The All Source Fusion Report

The CIA's most senior product is the all-source fusion report.

It takes raw intelligence from every collection discipline OSINT, HUMINT, signals intelligence, imagery and fuses it into a single coherent picture.

The fusion is where individual signals that mean nothing alone combine into conclusions that are actionable.

"Run a full all-source fusion on [competitor / market / situation]. Take everything collected across signals collection, OSINT, pattern of life, and human source analysis and fuse it into one coherent intelligence picture. Where do multiple independent sources confirm the same conclusion these are your high-confidence findings. Where do sources contradict each other these are your intelligence gaps requiring more collection. Where does one strong signal stand alone without corroboratio these require cautious treatment. Produce a fused assessment with every conclusion rated by confidence level: high, medium, or low based on the quality and quantity of sourcing behind it."

One source that says something is interesting.

Five independent sources that say the same thing is intelligence.Image
9/ The Competitive Counterintelligence Audit

The CIA doesn't just collect intelligence on others.

They run counterintelligence operations to identify what others are collecting on them.

Most companies never ask this question.

What does our competition already know about us and how are they finding it?

"Run a counterintelligence audit on [your company]. From the perspective of a sophisticated competitor running the same intelligence collection methods against us: what would they already know from our public signals? What does our job posting history reveal about our strategic priorities and capability gaps? What do our employees reveal on LinkedIn, Glassdoor, and social media about our internal operations? What does our pricing behavior, product release cadence, and partnership activity telegraph about our next moves? What intelligence gaps do we have that a sophisticated competitor could exploit? What are we broadcasting that we should stop broadcasting?"

The best defense against competitive intelligence is knowing what you're already revealing.Image
10/ The Strategic Warning Assessment

CIA strategic warning is the highest-stakes intelligence product.

It answers one question: is a fundamental shift coming that will change the entire operating environment?

Not a tactical threat. A strategic one.

The kind of shift that makes your current strategy obsolete before you realize it's happened.

"Produce a strategic warning assessment for [your industry]. Look beyond the next 6 months. What are the 3 scenarios that would fundamentally restructure this industry in the next 3 to 5 years the changes that would make the current competitive map irrelevant? For each scenario: what is the current probability, what are the early warning indicators already visible, how much warning time would we realistically have before the shift becomes irreversible, and what strategic position would be most advantaged versus most vulnerable if this scenario materializes? Which scenario is most underpriced by the market right now?"

Most strategy is optimized for the world as it is.

Strategic warning tells you when to start optimizing for the world as it's about to be.Image
11/ The Intelligence Gap Analysis

CIA analysts are trained to be as rigorous about what they don't know as what they do.

Every finished intelligence product includes a formal gaps assessment the questions that collection hasn't answered, the uncertainties that remain, the places where confidence should be lower than the analysis might suggest.

Most business intelligence skips this entirely.

Which is how companies end up acting on incomplete pictures with full confidence.

"Run a formal intelligence gaps analysis on everything collected about [competitor / market / situation]. For every major conclusion in the analysis: what would we need to know to be genuinely confident in this finding versus cautiously confident? What questions remain unanswered that could materially change the assessment if answered differently? What are we assuming in the absence of direct evidence and how sensitive is the overall conclusion to those assumptions being wrong? Produce a ranked list of priority intelligence requirements: the specific questions that if answered would most improve the accuracy and confidence of the overall assessment."

Knowing what you don't know is half of knowing what you do.Image
12/ The Daily Intelligence Brief

The President's Daily Brief is the most tightly edited document in the world.

One page. Maximum. Written for a reader with no time and every decision pending.

Every word earns its place or it doesn't appear.

It doesn't summarize everything collected. It tells the decision-maker the one thing they need to know today and what it means for the decisions they're making this week.

Run this every Monday morning on your competitive situation:

"Produce a one-page competitive intelligence daily brief for [your company and competitive situation]. Structure it exactly as the CIA structures the PDB: the most significant development from the last 7 days and why it matters now, the one trend that changed or accelerated this week that I need to factor into decisions, the one threat that moved closer to materialization and what the updated warning timeline looks like, and the single most important action or decision this intelligence suggests I should take before next Monday's brief. Maximum 300 words. No padding. Write for a decision-maker who will read this in 90 seconds and needs to walk away knowing what to do differently today."

The CIA brief doesn't inform.

It drives decisions.

That's the only standard worth building to.Image
The CIA's advantage was never classified information.

It was a disciplined process for turning public signals into actionable intelligence that decision-makers could act on with confidence.

Every method in these 12 prompts is built on that process.

Collection before analysis.
Multiple sources before conclusions.
Confidence ratings before recommendations.
Warning systems before events.

Most companies react to competitive threats after they've already materialized.

The ones that see them coming are running some version of this process.

Now you are too.

12 prompts. One Perplexity session.

The same competitive intelligence capability governments pay millions to maintain.

Save this thread.

Run prompt 1 on your most dangerous competitor this week.

By prompt 12 you'll know more about their next move than they think you do.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Spencer Baggins

Spencer Baggins Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @bigaiguy

Apr 13
Claude can now build trading algorithms like Goldman Sach's algorithmic trading desk (for 100% free).

Here are 15 insane Claude prompts that replace $500K/year quant strats (Save for later) Image
1. The Goldman Sachs Quant Strategy Architect

"You are a managing director on Goldman Sachs' algorithmic trading desk who designs systematic trading strategies managing $10B+ in institutional capital across global equity markets.

I need a complete quantitative trading strategy designed from scratch.

Architect:

- Strategy thesis: the specific market inefficiency or pattern this strategy exploits
- Universe selection: which instruments to trade and why (stocks, ETFs, futures, options)
- Signal generation logic: the exact mathematical rules that produce buy and sell signals
- Entry rules: precise conditions that must all be true before opening a position
- Exit rules: profit targets, stop losses, time-based exits, and signal reversal exits
- Position sizing model: how much capital to allocate per trade based on conviction and risk
- Risk parameters: maximum drawdown, position limits, sector exposure caps, and correlation limits
- Backtesting framework: how to properly test this strategy against historical data
- Benchmark selection: what to measure performance against and why
- Edge decay monitoring: how to detect when the strategy stops working

Format as a Goldman Sachs-style quantitative strategy memo with mathematical formulas, pseudocode logic, and risk parameter tables.

My trading focus: [DESCRIBE YOUR CAPITAL, PREFERRED MARKETS, TIME HORIZON, RISK TOLERANCE, AND ANY STRATEGIES YOU'VE EXPLORED]"
2. The Renaissance Technologies Backtesting Engine

"You are a senior quantitative researcher at Renaissance Technologies who builds rigorous backtesting systems that separate real alpha from overfitted noise across decades of market data.

I need a complete backtesting framework that gives me honest, reliable results.

Build:

- Data requirements: which historical data feeds I need, minimum time periods, and data quality checks
- Backtesting engine architecture: event-driven or vectorized with pros and cons for my strategy type
- Transaction cost modeling: commissions, slippage, bid-ask spread, and market impact estimates
- Lookahead bias prevention: safeguards that ensure no future data leaks into past decisions
- Survivorship bias handling: accounting for delisted stocks and failed companies in historical data
- Walk-forward optimization: train on past data, test on unseen data in rolling windows
- Out-of-sample testing protocol: how to split data so results aren't just curve-fitting
- Monte Carlo simulation: randomize trade sequences to understand the range of possible outcomes
- Statistical significance tests: is the backtest return real or could it happen by random chance
- Complete Python backtesting code ready to run with sample data and visualization

Format as a quantitative research document with full Python code, statistical validation methodology, and result interpretation guidelines.

My strategy: [DESCRIBE YOUR TRADING STRATEGY, PREFERRED MARKET, TIME FRAME, AND AVAILABLE HISTORICAL DATA]"
Read 17 tweets
Apr 11
🚨 BREAKING: AI can now teach machine learning like Stanford's CS229 professors (for free).

Here are 15 insane Claude prompts that replace $120,000 ML bootcamps (Save for later) Image
1. The Stanford CS229 Learning Roadmap Builder

"You are a professor at Stanford who teaches CS229 (Machine Learning) and has guided thousands of students from zero ML knowledge to landing $300K+ jobs at Google Brain, DeepMind, and OpenAI.

I need a complete personalized machine learning study plan based on my current skill level.

Build:

- Skill assessment: test my current knowledge and identify exact gaps to fill
- Learning path: week-by-week curriculum from my starting point to my target ML role
- Math prerequisites: exactly which linear algebra, calculus, probability, and statistics topics I actually need
- Resource curation: the single best free resource for each topic (no overwhelming lists of 50 links)
- Project milestones: a hands-on project at the end of each phase that proves I learned the concept
- Tool setup: exactly what to install (Python, Jupyter, scikit-learn, PyTorch) with setup instructions
- Time estimate: realistic hours per week needed and total months to reach my goal
- Common traps: mistakes self-learners make that waste months and how to avoid each one
- Portfolio plan: 5 projects that prove ML competence to hiring managers
- Interview readiness checklist: what I need to know to pass ML interviews at top tech companies

Format as a Stanford-style course syllabus with weekly topics, assignments, readings, and milestone checkpoints.

My background: [DESCRIBE YOUR CURRENT CODING SKILL, MATH LEVEL, ML EXPERIENCE, AVAILABLE HOURS PER WEEK, AND CAREER GOAL]"
2. The Andrew Ng Math-to-Intuition Translator

"You are Andrew Ng teaching machine learning at Stanford, famous for making complex math feel like common sense by using simple analogies, visual explanations, and real-world examples that anyone can understand.

I need a specific ML concept explained so clearly that a smart 12-year-old could understand it.

Explain:

- One-sentence summary: what this concept does in plain everyday language
- Real-world analogy: compare it to something from daily life that works the same way
- Visual description: paint a mental picture I can see in my head without any equations
- Why it matters: what problem this concept solves and what would happen without it
- The math behind it: equations introduced gently, one piece at a time, with every symbol explained
- Worked example: walk through a tiny numerical example by hand, step by step
- Python code: 10-15 lines of code that implement this concept from scratch (no libraries hiding the logic)
- Common confusions: the top 3 misunderstandings beginners have and the correct way to think about it
- Connection to other concepts: how this relates to things I already know
- One quiz question: test whether I actually understood it with the answer explained

Format as an Andrew Ng-style lecture note with the analogy first, math second, and code third.

The concept I want explained: [NAME THE ML CONCEPT AND DESCRIBE YOUR CURRENT UNDERSTANDING LEVEL]"
Read 18 tweets
Apr 6
I scraped every single NotebookLM prompt that blew up on X, Reddit, and academic corners of the internet.

Turns out most people are using NotebookLM like a fancy note-taker.

That's insane.

It's a full-blown research assistant that can compress 10 hours of analysis into 20 seconds if you feed it the right instructions.

Here's what actually works:Image
Prompt 1: The Expert Synthesizer

"You are a [field] expert with 15 years of experience. Analyze these sources and identify the 3 core insights that practitioners in this field would immediately recognize as groundbreaking. For each insight, explain why it matters and what conventional wisdom it challenges."

This forces depth over breadth. The output is immediately usable.
Prompt 2: The Contradiction Hunter

"Compare these sources and identify every point where they contradict each other. For each contradiction, explain which source has stronger evidence and why. If both are credible, explain what factors might explain the disagreement."

Perfect for literature reviews and due diligence. Saves hours of manual cross-referencing.
Read 14 tweets
Mar 21
OpenAI, Anthropic, and Google AI engineers use 10 internal prompting techniques that guarantee near-perfect accuracy…and nobody outside the labs is supposed to know them.

Here are 10 of them (Save this for later): Image
Technique 1: Role-Based Constraint Prompting

The expert don't just ask AI to "write code." They assign expert roles with specific constraints.

Template:

You are a [specific role] with [X years] experience in [domain].
Your task: [specific task]
Constraints: [list 3-5 specific limitations]
Output format: [exact format needed]

---

Example:

You are a senior Python engineer with 10 years in data pipeline optimization.
Your task: Build a real-time ETL pipeline for 10M records/hour
Constraints:
- Must use Apache Kafka
- Maximum 2GB memory footprint
- Sub-100ms latency
- Zero data loss tolerance
Output format: Production-ready code with inline documentation

---

This gets you 10x more specific outputs than "write me an ETL pipeline."

Watch the OpenAI demo of GPT-5 and see how they were prompting ChatGPT... you will get the idea.
Technique 2: Chain-of-Verification (CoVe)

Google's research team uses this to eliminate hallucinations.

The model generates an answer, then generates verification questions, answers them, and refines the original response.

Template:

Task: [your question]

Step 1: Provide your initial answer
Step 2: Generate 5 verification questions that would expose errors in your answer
Step 3: Answer each verification question
Step 4: Provide your final, corrected answer based on verification

---

Example:

Task: Explain how transformers handle long-context windows

Step 1: Provide your initial answer
Step 2: Generate 5 verification questions that would expose errors in your answer
Step 3: Answer each verification question
Step 4: Provide your final, corrected answer based on verification

---

Accuracy jumps from 60% to 92% on complex technical queries.Image
Read 13 tweets
Mar 19
BREAKING: Stanford researchers just published a prompting technique that makes today’s LLMs behave like better versions of themselves.

It’s called “prompt ensembling” and it runs 5 variations of the same prompt, then merges the outputs.

Here’s how it works 👇 Image
The concept is simple:

Instead of asking your question once and hoping for the best, you ask it 5 different ways and combine the answers.

Think of it like getting second opinions from 5 doctors instead of trusting one diagnosis.

Stanford tested this on GPT-5.2, Claude 4.5, and Gemini 3.0.Image
Here's the systematic approach to generate prompt variations:

Variation 1: Direct Question
"What are the main risks of [decision]?"

Variation 2: Role-Based
"You're a risk analyst with 20 years experience. What are the main risks of [decision]?"

Variation 3: Contrarian Angle
"What would a critic say are the main risks of [decision] that most people overlook?"

Variation 4: First Principles
"Breaking this down from first principles, what are the fundamental risks of [decision]?"

Variation 5: Historical Comparison
"Based on similar decisions in the past, what are the main risks of [decision]?"

Each variation triggers different reasoning paths in the LLM.

Now comes the merging strategy...Image
Read 12 tweets
Mar 16
BREAKING: Perplexity can now do competitive intelligence like Deloitte, PwC, and McKinsey.

Here are 14 prompts that replace $120K/year corporate strategy work (Save this thread) Image
1/ Full Competitive Landscape Map

"You are a Senior Strategy Consultant at McKinsey. I need a complete competitive landscape for [COMPANY NAME].

Use live web data and cite sources.

Please provide:
- Top 15 direct competitors
- 10 indirect/substitute competitors
- Market share estimates
- Revenue comparison table
- Geographic presence comparison
- Business model differences
- Positioning summary (premium, mid-market, low-cost)
- Visual 2x2 strategic map description

Format as a board-ready competitive landscape slide.

Company: [INSERT COMPANY + INDUSTRY]"
2/ Institutional SWOT Analysis

"You are a Strategy Partner at Bain. I need a deep SWOT analysis for [COMPANY NAME] using recent data.

Please provide:
- Strengths with evidence (financials, brand, tech)
- Weaknesses backed by metrics
- Opportunities tied to industry trends
- Threats including regulatory and macro risks
- Which quadrant matters most right now
- Strategic priorities based on SWOT

Format as executive strategy memo.

Company: [INSERT COMPANY]"
Read 17 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(