Don’t Invest Until You Run These 3 Bear Case Prompts

How to destroy your own stock idea before the market does it for you

Hello, fellow stock pickers

Charlie Munger had one rule before forming an opinion:

“I never allow myself to hold an opinion on anything that I don’t know the other side’s argument better than they do.”

In investing, this isn’t optional.
Always know the other side before you act.

When I started investing, I broke this rule sometimes.
I’d get excited about a stock.
And skip the hard part : finding reasons I might be dead wrong.

Long term investing isn’t about finding winners.
It’s about avoiding the losers that wipe out years of compounding.

In 2025, we have the perfect sparring partner for this: AI.
It never gets tired, never gets emotional, and will happily rip your idea to shreds in minutes.

Here’s the exact 3 prompt sequence I run on every stock before I put money in.

Let’s dive in:

Prompt 1: Run a “Short-Seller Report”

Before you buy a stock, flip sides.
Pretend you’re a short-seller writing a hit piece on your own idea.

Here’s my process:

• Take the company (e.g., Planisware)

• Drop it into Gemini Deep Search (my go-to for deep analysis)

• Ask it to find every way this stock could blow up

• Go through each red flag, one by one

• Kill it with facts and evidence

• If I can’t kill it , I walk away

The danger zones I always check:

• Fraud / misconduct

• Aggressive accounting

• Business-model flaws

• Unrealistic growth

• Competitive & macro threats

• Governance issues

• Valuation

Here is the prompt I ran:

Target : Planisware SA (Euronext Paris )


---

 ROLE

Senior forensic short-seller (ex–Big Four forensic accountant) in professional asset management.


 OBJECTIVE

Surface and rank empirically verifiable vulnerabilities that could drive material share-price downside within 24 months.

*Skip any deep, theoretical teardown of the company’s business model or product strategy; focus strictly on red-flag evidence.*


---


 DATA-GATHERING — *trigger DeepSearch*


| Bucket | Must-pull docs (last 5 yrs) | Quick query hints |

|--------|-----------------------------|-------------------|

| **Core Filings** | URD / Doc. de Réf. (AMF), Infogreffe statutory accounts, BALO notices, Banque de France liens | `"Planisware" "rapport de gestion"` |

| **Transcripts** | Earnings webcasts, broker conferences | `"Planisware" "Q&A"` |

| **Alt-Data** | Glassdoor & LinkedIn attrition, INPI/EUIPO disputes, DGDDI trade data (HS 8523), SimilarWeb traffic, app-store ratings, AMF insider trades, EuroTLX short-interest | `"Planisware" "turnover" OR "départs"` |

| **Opinions** | Broker downgrades, ESG notes, activist posts | `"Planisware" "sell" OR "short thesis"` |


> **Save full URL / file path + page or line # for every excerpt.**


---


 2 ANALYSIS BUCKETS (score each finding **Severity 1-5 × Probability 1-5**)

A Accounting/Fraud B Capital Allocation/Liquidity C Governance & Incentives D Customer/Supplier Risk E Macro Catalysts


*(No extended business-model section—only include if a fact-based weakness surfaces in the above buckets.)*


---


 3 REPORT FRAME (flex length)


1. **Executive Summary** ≤ 1 page

• Headline downside range + three one-line “kill-shot” theses.


2. **Ranked Issue Table** (Top 10-15)

| Rank | Bucket | Evidence (File + Pg) | Severity | Prob. | 12-mo Px-Impact % | One-liner |


3. Detailed Findings

One sub-section per bucket; numbered footnotes link to sources.


4. Falsifiability Checks

What could disprove each core bear point.


5. Missing-Data Flags

Critical gaps + follow-up query strings.


---


 4 STYLE RULES

* Cite exact file + page for **every** claim.

* No adjectives without numbers.

* Bullets > prose; zero filler.

* Tables first, narrative second.


---


 5 DELIVERY

Single **.docx or PDF** dossier. No CSVs, slides, or appendices unless requested. 

The results were great.
It flagged risks I’d never even considered.

Yes, it’s too negative… that’s the point.
It’s there to cool down the rush of a new idea before you put money in.

Here’s what to do:

• Take the prompt (make it yours )

• Drop in your company’s name

• Run it in a Deep Search (not normal ChatGPT mode )

What came back for me?
A 10-page short-seller take that opened my eyes to a few blind spots.

Planisware Short Sell Forensic Dossier

237KB ∙ PDF file

Download

Download

Prompt 2: The Product Reality Check

Prompt 1 looks at the business model.
This one looks at what customers really think.

Use AI to find out why customers will stop using the products (e.g., Planisware’s software ) or choose a competitor instead.

How to do it:

• If it’s a B2B company, check professional review sites and industry reports.

• If it’s a B2C company, look at reliable consumer review sites and verified feedback.

• Find the 5 main reasons customers leave.

• Compare the company to its top 3–4 rivals

• List the key weak spots where the company is losing ground.

• Finish with 5 risks that could shrink revenue or weaken the moat.

Run this check before you invest.
If customers are already walking out the door, the numbers will catch up.
Here’s the prompt , steal it as always and make it yours.

ROLE: You are a market intelligence analyst for a long-term investor, tasked with identifying potential weaknesses in [Company] based on verified customer data.

TASK:
1. From verified sources (Gartner Peer Insights, G2, TrustRadius, public RFP docs, competitor marketing):
   - Identify the top 5 SWITCH TRIGGERS that cause customers to leave or choose competitors. Provide 2+ direct quotes for each, with source/date.
   - Benchmark [Company] vs [Competitor 1], [Competitor 2], [Competitor 3] on Price, Performance, Support, Brand:
       • Score each 1–10.
       • Include 1 quote + source/date per score.
       • If no verified data, state “No verified data”.
2. Highlight comparative weaknesses revealed in the benchmark.
3. Summarize 3–5 key investment risks (moat erosion, churn risk, pricing pressure).

RULES:
- All claims must be supported by verifiable buyer quotes; no speculation.
- Prioritize reviews dated within [Timeframe].
- Keep findings specific, avoiding generic “poor UX” or “too expensive” without context.

OUTPUT:
Section 1: Switch Triggers (ranked, quotes inline).
Section 2: Competitor Benchmark Table (4 attrs × 4 vendors).
Section 3: Comparative Weaknesses (bullets).
Section 4: Investment Risk Summary (bullets).

Prompt 3: The Thesis Breaker

If Prompt 1 kills bad business models and Prompt 2 exposes unhappy customers, Prompt 3 is where you try to blow up your own idea.

The goal: take the thesis you believe in… and see if it survives an attack from you.

Here’s how I do it:

1. Write down the 5–10 pillars holding your thesis up.

2. Grab the Master Prompt below.

3. Paste it into ChatGPT (normal mode, no DeepSearch) it will adapt the prompt to your specific business.

4. Take that adapted prompt and drop it into DeepSearch.

5. Watch it come back with every weakness it finds.

   

Here’s the Master Prompt I use:

You are a senior buy-side analyst tasked with falsifying a long thesis for a public company.
I will give you:

    Company name

    Ticker / ISIN

Your job:

    Take the generic Thesis Breaker framework below.

    Adapt it to the company’s specific industry, competitors, and business model.

    Output a DeepSearch-ready prompt — ready to paste into Gemini DeepSearch or any research tool with web/doc access.

    Do not run the analysis — only return the adapted prompt.

GENERIC FRAMEWORK

ROLE
Senior buy-side analyst tasked with falsifying a long thesis.

INPUT

    Bull-thesis text — adapt it to the company’s real business model, market, and competitive positioning.

    Ticker / ISIN: [INSERT TICKER]

OBJECTIVE
Surface all empirically testable weak points that could break the thesis within 24 months.

TASKS
A. Parse the thesis → list max 10 explicit or implied assumptions.
Examples:

    Revenue growth target

    Retention rates

    Recurring revenue share

    Margins

    Management alignment

    Competition risk

    Churn limits

    FX impact

    Regulatory/tax impact

    TAM growth

B. For each assumption:

    Classify as Fact / Projection / Opinion

    Pull contrary evidence from:
    • Official filings (10-K, annual reports, URD, prospectus)
    • Earnings call transcripts
    • Investor presentations
    • Industry reports
    • Alternative data (traffic, job postings, Glassdoor, customer reviews)

    Cite exact source + page/line/date

    Rate Fragility (1 low → 5 high) and Impact (1 low → 5 high share-price hit)

C. Rank the top 5 “kill-switch” assumptions by Fragility × Impact.

D. Flag missing data that blocks a verdict.

E. Provide a 1-page summary:

    Headline downside range

    Three biggest kill-shots

STYLE RULES

    Bullet tables only

    No adjectives without numbers

    Every claim must have a cited source

    Max 5 evidence bullets per assumption

OUTPUT

    Ranked Kill-Switch Table (Fragility × Impact, % downside)

    Evidence bullets under each assumption

    Missing Data list

Now adapt this framework to:

    Company: [INSERT COMPANY NAME]

    Ticker/ISIN: [INSERT TICKER OR ISIN]

Make sure the adapted prompt names:

    The exact competitors in the same segment

    The specific metrics the company reports

    The realistic assumptions based on its latest filings and market context

---

Recap: Stress Test Any Stock in 3 Steps

• Run a Short-Seller Report
  Flip sides. Hunt for every red flag in the business model.

• Do a Product Reality Check
  Find out why customers might leave and how it stacks up against rivals.

• Break Your Own Thesis
  List your key assumptions and try to disprove them with facts.

It works for any stock.
And once your prompts are ready, you can reuse them forever.

3 prompts.
1 process.
A full bear-case check without spending weeks on research.