What Is the Mirror Test in AI Berkshire’s Investment Decision-Making?

The Mirror Test is a mandatory five-sentence sanity check that forces analysts to articulate the business model, moat, management, valuation, and downside risk before any investment is approved.

In the AI Berkshire workflow, the Mirror Test (镜子测试) serves as a non-negotiable quality gate. Every company must pass this disciplined review before advancing to deeper research or final investment recommendations. According to the repository’s core policy documented in README.md, if an analyst cannot complete the five-sentence template, the decision is automatically 未通过 (not passed) and the stock is categorically excluded from purchase.

The Five Pillars of the Mirror Test

The Mirror Test requires a concise statement covering five specific pillars. Each pillar maps to a single sentence in the output template.

1. Business Model – A clear explanation of what the company does and why the analyst understands the mechanics of the business.

2. Moat (护城河) – The competitive advantage and whether it is widening or eroding over time.

3. Management Team – An assessment of whether leadership can be trusted with capital allocation and shareholder interests.

4. Valuation / Safety Margin – The current price relative to intrinsic value, expressed as a discount (e.g., “8‑fold discount” or “8 折”).

5. Downside Risk – A concrete statement explaining why potential losses would be limited even if the investment thesis fails.

If any pillar is missing or unconvincing, the rule documented in skills/investment-checklist.md (lines 39‑40) triggers an immediate rejection: “5句话说不完整 = 不买” (Five sentences incomplete = do not buy).

How the Mirror Test Fits Into the Architecture

The Mirror Test is embedded across three architectural layers in the AI Berkshire system.

Skill Layer – The test is implemented in the /investment-checklist skill defined in skills/investment-checklist.md. This skill executes after six preliminary “关” (gates) checks, with the Mirror Test serving as the fourth step in the sequence.

Agent Layer – Four parallel agents (段永平, 巴菲特, 芒格, and 李录) each provide inputs that feed directly into the five‑sentence template. This multi‑disciplinary perspective ensures the final thesis incorporates diverse value‑investing philosophies before capital allocation.

Tool Layer – The completed five‑sentence block is parsed by downstream validation tools. Specifically, tools/financial_rigor.py verifies the valuation claim, while the thesis‑tracker monitors whether the articulated investment case holds over time.

Implementation in Source Code

The Mirror Test template is hardcoded in the investment checklist skill. Below is a Python implementation that replicates the logic found in skills/investment-checklist.md (lines 75‑84), allowing you to generate compliant Mirror Test statements programmatically.

def mirror_test(
    price: str,
    company: str,
    business: str,
    moat: str,
    management: str,
    valuation: str,
    downside: str,
) -> str:
    """
    Build a 5-sentence Mirror Test statement matching the AI Berkshire template.
    """
    tmpl = (
        f'我以 {price} 买入 {company}，因为：\n'
        f'1. 这门生意的本质是 {business}，我理解它；\n'
        f'2. 它的护城河是 {moat}，而且在变宽/变窄；\n'
        f'3. 管理层 {management}，值得/不值得信赖；\n'
        f'4. 当前价格相当于内在价值的 {valuation} 折，有/无足够安全边际；\n'
        f'5. 即使我错了，下行风险 {downside}，因为 {downside}。'
    )
    return tmpl

# Example usage replicating a real entry from the repository:

print(
    mirror_test(
        price="380港元",
        company="腾讯",
        business="社交网络+数字内容平台",
        moat="12亿用户的社交关系链",
        management="Pony Ma 低调务实、资本配置优秀",
        valuation="8",
        downside="可控，因为账上净现金超2000亿、游戏现金流强劲"
    )
)

Output:

我以 380港元 买入 腾讯，因为：
1. 这门生意的本质是 社交网络+数字内容平台，我理解它；
2. 它的护城河是 12亿用户的社交关系链，而且在变宽/变窄；
3. 管理层 Pony Ma 低调务实、资本配置优秀，值得/不值得信赖；
4. 当前价格相当于内在价值的 8 折，有/无足够安全边际；
5. 即使我错了，下行风险 可控，因为账上净现金超2000亿、游戏现金流强劲。

This format matches the live examples found in reports/巴菲特镜子测试-7家公司-20260419.md (lines 41‑48), where the five‑sentence block determines the final pass/fail status for each company under review.

Enforcement and the "未通过" Logic

The Mirror Test operates as a hard stop with no exceptions allowed.

• Pass Condition: All five sentences are present and substantively address the required pillars.
• Fail Condition: Any missing or vague sentence triggers the 未通过 status.

As documented in README.md (lines 73‑75) and reinforced in codex-skills/investment-checklist/SKILL.md (lines 189‑191), this logic is absolute. The system treats incomplete theses as automatic disqualifications, preventing capital allocation until the analyst can articulate a complete, coherent investment case.

Summary

• The Mirror Test is a mandatory five‑sentence sanity check in the AI Berkshire investment workflow.
• It requires explicit statements on business model, moat, management, valuation, and downside risk.
• The rule “5句话说不完整 = 不买” creates a hard stop that prevents incomplete theses from advancing.
• Implementation spans skills/investment-checklist.md, parallel agent inputs, and downstream validation in tools/financial_rigor.py.
• Real‑world examples in the repository demonstrate the template’s application to specific securities like Tencent.

Frequently Asked Questions

What happens if an analyst cannot complete all five sentences in the Mirror Test?

If any of the five required sentences are missing or unconvincing, the investment receives a 未通过 (not passed) status. According to the logic in skills/investment-checklist.md (lines 39‑40), the stock is automatically rejected and cannot be purchased until a complete thesis is articulated.

Where is the Mirror Test template defined in the AI Berkshire codebase?

The primary template is defined in skills/investment-checklist.md (lines 75‑84). A mirrored version for Codex integration appears in codex-skills/investment-checklist/SKILL.md (lines 189‑191). Both files enforce the same five‑sentence structure and rejection logic.

How does the Mirror Test interact with the four parallel agents?

The four agents (段永平, 巴菲特, 芒格, and 李录) each generate inputs that populate the five‑sentence template. This ensures the final Mirror Test statement reflects a multi‑disciplinary value‑investing perspective before undergoing downstream validation by tools/financial_rigor.py.

Can the Mirror Test be bypassed for urgent investment opportunities?

No. The Mirror Test is treated as a disciplinary gate with no exceptions permitted. The repository’s policy in README.md (lines 73‑75) explicitly states that incomplete five‑sentence explanations result in automatic rejection, regardless of time constraints or market conditions.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}