# Build an investment-thesis file from expert commentary

> Collect the strongest bull and bear cases into a library, then stress-test your thesis against all of them in one chat.

- For: Investor
- Category: Finance, Investing & Due Diligence
- MCP tools: deep_query, library_import_youtube_video, deep_context
- HTML version: https://veridive.com/mcp/use-cases/investment-thesis-file

## The problem

A thesis you've only defended against your own doubts is undefended. The failure mode is universal: you form a view, and from that moment every video you click agrees with you — the algorithm makes sure of it. The bear case exists, articulated brilliantly by someone somewhere, and you'll meet it for the first time in the drawdown.

The fix is structural, not motivational: build the opposition a permanent seat. A thesis *file* — the strongest arguments on both sides, collected deliberately, kept current, and interrogated as a set — turns "I believe X" into "I believe X, having heard the best versions of not-X, and here's where they're wrong or where I've sized the risk."

## How veridive MCP helps

**deep_query** finds the file's contents without the algorithm's help: ask for the strongest bull case and the strongest bear case on the asset or theme, and it researches across the spoken record — fund managers, operators, short sellers, domain experts — returning both sides **with citations**. You're sampling the argument space, not your recommendation feed.

**library_import_youtube_video** pins the keepers: the ten or fifteen pieces that argue best get imported, tagged bull/bear, transcripts and all. **deep_context** then makes the file argue with you: scope a question across the collection — *"which bear arguments does my thesis not answer?"* — and get a grounded response citing the exact chunks. The file becomes a standing sparring partner that updates as you add to it.

## How to run it

1. **Connect veridive MCP** — One-time setup: add the veridive MCP server to Claude, ChatGPT, Cursor or any MCP client and generate your API key. Takes about two minutes.
2. **Sample both sides deliberately** — Ask for the best of each camp — explicitly. This is the anti-algorithm step: you're requesting the case you wouldn't be shown.

   ```
   What are the strongest bull arguments and the strongest bear arguments on <asset/theme>, as made by credible people on the record? Both sides, cited — use veridive's deep_query.
   ```
3. **Pin the keepers into the file** — Import the videos that argue best, tagged by side. The file is small on purpose — strongest versions only.

   ```
   Import these into my veridive library, tagged '<theme>-bull' or '<theme>-bear' accordingly: <URLs>.
   ```
4. **Make the file argue with you** — Stress-test against the whole collection. The citations tell you exactly which argument, from whom, your thesis hasn't answered.

   ```
   Here's my thesis: <thesis>. Against the tagged '<theme>' videos in my library: which arguments does it fail to address, and which does it survive? Cite the chunks.
   ```

## Problems it solves

- Recommendation algorithms turn a view into an echo chamber within a week of clicks.
- The best bear case is scattered across channels you'll never be shown while bullish.
- Thesis reviews test against your own summary of the opposition — a strawman by construction.
- Six months in, you can't remember which risks you consciously accepted versus never heard.

## Why it's valuable

- The opposition gets a permanent, well-argued seat — dissonance by design, before the market provides it.
- Stress-tests cite the exact argument and moment, so gaps in the thesis are specific and fixable.
- The file compounds: each earnings cycle's best new arguments join, and the same questions re-run.
- Sizing improves when accepted risks are written down with sources instead of vaguely sensed.

## Example prompts

```
Strongest bull and bear cases on <asset>, from credible on-the-record sources, cited — use veridive's deep_query.
```

```
Import these thesis-file videos, tagged '<theme>-bull'/'<theme>-bear': <URLs>.
```

```
My thesis: <thesis>. Which arguments in my '<theme>' file does it not answer? Cite chunks.
```

## FAQ

**How big should the file be?**

Small and sharp — the ten to fifteen strongest pieces, not everything on the topic. deep_context scopes a question across up to 20 items at a time, which is exactly the right size for a thesis file anyway.

**Is this investment advice?**

No. It's an argument-collection and stress-testing workflow over public commentary. What you conclude, buy or size is entirely your judgment — the tool just guarantees you heard the other side first.

**How do I keep the file current?**

After big prints or narrative shifts, re-run the both-sides sweep and import anything that argues something new. Stale files fail silently — the re-run is the maintenance.

## Related use cases

- [Separate evidence from hype in a hot sector](https://veridive.com/mcp/use-cases/evidence-vs-hype.md) — Quality verdicts flag which videos back claims with data and which just vibe — filter a noisy sector down to the sourced few.
- [Compare rival experts on one question](https://veridive.com/mcp/use-cases/rival-experts-one-question.md) — "Is X dead?" — get the disagreement mapped: who says yes, who says no, and the timestamped why.
- [Draft the quarterly "what changed" memo](https://veridive.com/mcp/use-cases/quarterly-what-changed-memo.md) — Ground the quarter's shifts in imported commentary, then have chat shape it into a memo your PM can forward.

---

Part of the veridive MCP use-case library: https://veridive.com/mcp/use-cases.md
Server docs for agents: https://veridive.com/mcp.md