# AI Berkshire Information Richness Rating System Explained: A/B/C Levels Guide

> Understand the AI Berkshire information richness rating system (A/B/C levels). Learn how data volume dictates AI limitations disclosure for investment reports. Explore the xbtlin/ai-berkshire repository.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: deep-dive
- Published: 2026-06-26

---

**The AI Berkshire information richness rating system categorizes companies into A, B, or C levels based on publicly available data volume, directly determining research methodology and AI limitation disclosures required in every investment report.**

The `xbtlin/ai-berkshire` repository implements a structured framework for AI-driven investment research that classifies target companies by data availability. This rating system ensures that analysis depth and confidence levels match the underlying information quality, with specific implementation requirements defined in the skill configuration files located at [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) and [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md).

## What is the Information Richness Rating System?

The information richness rating system (信息丰富度评级) is a three-tier classification framework that quantifies the completeness and accessibility of public data for any given company. According to the source code in [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md), this rating serves as a critical determinant for research strategy, report structure, and the mandatory **AI research limitations statement** that must appear at the beginning of every generated report.

The system operates on a simple principle: the volume of available information dictates the AI's analytical approach. When information is abundant (A Level), the AI must actively guard against consensus bias; when information is scarce (C Level), the AI must explicitly flag uncertainty and estimation-based conclusions.

## The Three Rating Levels (A, B, and C)

### A Level – Information Abundant (信息充裕)

**A Level** ratings apply to established public companies with comprehensive data coverage. These entities typically have years of listed history, extensive analyst coverage, complete financial reporting (annual, quarterly, and special reports), and substantial media attention.

**Characteristics:**
- Complete financial statements with multi-year history
- Broad analyst coverage from multiple brokerages
- Extensive news and industry report availability
- High market consensus visibility

**Examples from the codebase:** Tencent, Alibaba, Kweichow Moutai, and NVIDIA are explicitly classified as A Level in their respective reports. The Tencent final report (`reports/腾讯/最终报告-20260411.md`) notes: "信息丰富度：A级——腾讯是全球关注度最高的中国互联网公司，信息极充裕."

### B Level – Moderate Information (信息适中)

**B Level** ratings indicate partial data availability. These companies may have shorter listing histories, limited analyst coverage, or gaps in public financial disclosure that require supplementary estimation or industry benchmarking.

**Characteristics:**
- Limited listed history or recent IPO
- Partial financial data requiring industry comparisons
- Moderate media and research coverage
- Some data gaps requiring AI estimation

**Examples:** Ant Group and Pop Mart (泡泡玛特) are categorized as B Level, reflecting their relatively shorter public histories or complex corporate structures that limit absolute data transparency.

### C Level – Information Scarcity (信息稀缺)

**C Level** ratings apply to private companies, early-stage startups, or entities with minimal public disclosure. Research on these targets relies heavily on estimated data, funding announcements, and limited media reports, requiring explicit uncertainty disclaimers.

**Characteristics:**
- No public listing or limited disclosure obligations
- Financial data derived from estimates or unofficial sources
- Limited or no analyst coverage
- High reliance on AI inference with low confidence intervals

**Examples:** Ant Digital (蚂蚁数科) and Lingjing Wanwei (灵境万维) are classified as C Level. The Lingjing Wanwei report (`reports/灵境万维-team-20260512/灵境万维_投资研究报告_20260512.md`) states: "信息丰富度：C级（信息稀缺）——灵境万维为未上市创业公司，信息稀缺."

## Impact on Research Strategy and AI Limitations

The rating system directly dictates two critical report components: the **research strategy** and the **AI limitations statement**.

**Research Strategy Adjustments:**
- **A Level:** Requires "反面检验" (devil's advocate testing) and non-consensus perspectives to avoid "correct but useless" consensus replication
- **B Level:** Balances available data with cautionary notes on estimation reliability
- **C Level:** Emphasizes uncertainty quantification and clearly labels all inferred data

As specified in [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md), every report must begin with both the rating classification and the corresponding limitations statement. This requirement ensures readers understand the confidence level and methodological constraints before proceeding with the analysis.

## Implementation in Report Templates

The [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) file mandates a standardized header format for all generated reports. Below is the implementation pattern showing how the rating and limitations statement appear in the final Markdown output:

```markdown

# AI Berkshire 研究报告 – {Company Name}

**信息丰富度评级：A 级（信息充裕）**  
AI 研究局限性声明：信息极度充裕易导致 AI 与市场共识趋同，报告重点放在反面检验和非共识视角，以避免“正确的废话”。

---

## 1. 商业模式分析（段永平视角）

...

```

For C Level companies, the limitations statement changes to reflect data scarcity:

```markdown
**信息丰富度评级：C 级（信息稀缺）**  
AI 研究局限性声明：本报告基于有限公开信息和 AI 推算数据，财务数据多为估算，不确定性较高，建议阅读者谨慎参考。

```

The [`skills/portfolio-review.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/portfolio-review.md) configuration also leverages these ratings during portfolio composition analysis, tagging each holding with its corresponding richness level to ensure weighted risk assessment across the investment universe.

## Summary

- **The AI Berkshire information richness rating system** classifies companies into A, B, or C levels based on data availability and public disclosure completeness.
- **A Level** (information abundant) applies to major listed companies like Tencent and Alibaba, requiring consensus-challenging research strategies.
- **B Level** (moderate information) covers companies with partial data requiring balanced estimation approaches.
- **C Level** (information scarcity) applies to private companies like Ant Digital, mandating explicit uncertainty warnings and low-confidence data labeling.
- **Implementation requirements** in [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) and [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) mandate that every report header include both the rating and a corresponding AI limitations statement.

## Frequently Asked Questions

### What does A Level mean in the AI Berkshire rating system?

A Level indicates information abundance, typically assigned to well-established public companies with comprehensive financial disclosures and extensive analyst coverage. According to the source code in [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md), A Level reports must specifically include "反面检验" (reverse stress testing) to avoid AI consensus bias, as seen in the Tencent and Alibaba research templates.

### How does the information richness rating affect AI research limitations?

The rating directly determines the AI limitations statement required at the start of every report. For A Level companies, the AI warns against consensus replication; for C Level companies like those referenced in `reports/灵境万维-team-20260512/`, the AI explicitly states that data is "推算" (estimated) and carries high uncertainty, prompting readers to exercise caution.

### What is the difference between B Level and C Level ratings?

B Level companies possess moderate public information—typically shorter listing histories or limited coverage where some data requires estimation. C Level companies, however, operate with severe information scarcity (often pre-IPO or private), where nearly all financial metrics are AI-estimated rather than reported, necessitating stronger confidence warnings in the final output.

### Where is the rating system defined in the ai-berkshire repository?

The rating system is formally defined across three primary skill files: [`skills/investment-team.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-team.md) establishes the rating's purpose for research strategy; [`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md) mandates the header format and AI limitations statement; and [`skills/portfolio-review.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/portfolio-review.md) utilizes the ratings for portfolio-level risk assessment.