# Difference Between Industry-Research and Industry-Funnel Skills in AI-Berkshire > Discover the difference between industry-research and industry-funnel skills for AI development. Understand qualitative value-chain scanning versus quantitative market filtering to identify target companies. - Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire) - Tags: deep-dive - Published: 2026-07-25 --- **Industry‑Research performs a panoramic value‑chain scan of an entire industry using qualitative "Four Masters" analysis, while Industry‑Funnel runs a quantitative multi‑stage filter to narrow the market down to three specific target companies.** AI‑Berkshire implements its investment research workflow as discrete **skills** that users invoke via chat commands like `/industry-research` and `/industry-funnel`. Understanding the **difference between industry‑research and industry‑funnel skills** is essential for analysts who need to decide whether to map an entire sector or select specific securities from a crowded market. ## Architectural Goals and Design Philosophy Both skills belong to the *investment‑research* family defined in the `skills/` directory, but they serve distinct analytical purposes. ### Industry-Research: Panoramic Value-Chain Mapping The `industry-research` skill aims to produce a **panoramic value‑chain scan** of an industry. According to the specification in [`skills/industry-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-research.md), this skill builds an industry‑wide map and applies the "Four Masters" framework (段永平, 巴菲特, 芒格, 李录) to each segment of the chain. It emphasizes **qualitative segmentation** and macro‑level risk assessment using Munger checklists and civilization trend analysis. ### Industry-Funnel: Quantitative Multi-Stage Screening In contrast, the `industry-funnel` skill—defined in [`skills/industry-funnel.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-funnel.md)—implements a **tiered filter‑funnel** that narrows the entire market down to **three final target companies**. This skill emphasizes **quantitative hard screens** such as PE thresholds, ROE requirements, cash‑flow ratios, and debt‑to‑asset filters, complemented by structured moat scoring. ## Workflow Execution Comparison The operational pipelines differ significantly in their step‑by‑step execution. **Industry‑Research Workflow:** 1. Build a logical investment chain for the specified industry. 2. Generate a full‑chain diagram visualizing the value flow. 3. Scan every listed public company within the sector. 4. Perform "Four Masters" deep‑dives on each *segment‑leader*. 5. Complete a risk checklist, civilization trend analysis, and portfolio suggestions. **Industry‑Funnel Workflow:** 1. Define **three market‑wide entry pools** based on activity (活跃度), price change (涨幅), and market cap (市值). 2. Apply **five hard value‑investment criteria**: PE, ROE, cash flow (现金流), debt ratio (负债率), and competitive moat (护城河). 3. Conduct structured analysis for a maximum of ten companies passing the screen. 4. Execute "Four Masters" deep‑dives on the final three selections. 5. Output ETF alternatives, industry‑level positioning notes, and data‑audit steps. ## Output Artifacts and Report Formats Both skills generate markdown reports in the `reports/` directory, but with different granularities. - **Industry‑Research** generates an **industry‑level report** at `reports/{行业}-industry-{date}.md`, containing segment analyses, risk matrices, and portfolio allocations. - **Industry‑Funnel** generates a **company‑selection report** at `reports/{行业}-funnel-{date}.md`, listing filtered candidates, inclusion/exclusion rationale, and deep‑dives on the final three companies. ## Technical Implementation and Shared Infrastructure Both skills are defined as **canonical markdown files** under the `skills/` directory. The repository’s build script [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) converts each source file into a Codex‑compatible description located at [`codex-skills/industry-research/SKILL.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-skills/industry-research/SKILL.md) and [`codex-skills/industry-funnel/SKILL.md`](https://github.com/xbtlin/ai-berkshire/blob/main/codex-skills/industry-funnel/SKILL.md), enabling both Claude‑Code and Codex users to invoke the workflows. Shared infrastructure includes: - **[`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py)**: A utility used by both skills to verify data quality. - **[`skills/financial-data.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/financial-data.md)**: Provides data‑source guidelines leveraged by both pipelines. ## Practical Usage Examples Invoke these skills through the AI‑Berkshire chat interface or CLI: ```text # Map the entire nuclear power value chain /industry-research 核电 ``` *Result:* A markdown report `reports/核电-industry-YYYYMMDD.md` containing a sector‑wide chain diagram, segment‑by‑segment "Four Masters" analysis, and portfolio suggestions. ```text # Select top 3 AI computing stocks using quantitative filters /industry-funnel AI算力 ``` *Result:* A markdown report `reports/AI算力-funnel-YYYYMMDD.md` listing the initial 30‑60 candidates, the 5‑hard‑criteria screening results, structured analyses for the survivors, and a final three‑company deep‑dive. ## Summary - **Industry‑Research** conducts **macro‑level qualitative mapping** of entire value chains, ideal for understanding *where* value exists in an industry. - **Industry‑Funnel** executes **micro‑level quantitative filtering** to identify specific securities, ideal for determining *which* companies merit investment. - Both skills share common infrastructure like [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) and [`tools/report_audit.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/report_audit.py), but diverge in their primary output: industry reports versus company selection lists. - Analysts should typically run **Industry‑Research first** to understand the sector landscape, followed by **Industry‑Funnel** to select specific targets. ## Frequently Asked Questions ### Which skill should I run first when analyzing a new industry? Run **`industry‑research`** first. This skill provides the panoramic value‑chain scan necessary to understand segment dynamics and identify where value concentrates. Once you understand the industry structure, **`industry‑funnel`** can quantitatively filter for the best individual securities within that mapped landscape. ### What are the "Four Masters" criteria referenced in both skills? The "Four Masters" (段永平, 巴菲特, 芒格, 李录) represent a qualitative investment framework applied to deep‑dive analyses. While `industry‑research` applies this framework to every *segment* of the value chain, `industry‑funnel` applies it only to the final three companies surviving the quantitative screens, ensuring rigorous qualitative validation of quantitative winners. ### How do the generated reports differ between the two skills? `industry‑research` outputs `reports/{行业}-industry-{date}.md`, which contains an industry‑wide map, segment risk matrices, and broad portfolio allocations. `industry‑funnel` outputs `reports/{行业}-funnel-{date}.md`, which contains a filtered shortlist, hard‑criteria screening results, and deep‑dives on exactly three target companies. The former is strategic; the latter is tactical. ### Can I modify the hard criteria in Industry-Funnel? The five hard criteria (PE, ROE, cash flow, debt ratio, and moat) are defined in [`skills/industry-funnel.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/industry-funnel.md) as part of the skill specification. To modify these thresholds, you must edit the source markdown file or create a custom skill variant, then regenerate the Codex artifact using [`scripts/sync-codex-skills.py`](https://github.com/xbtlin/ai-berkshire/blob/main/scripts/sync-codex-skills.py) to apply the changes.