# AI Berkshire Performance Track Records: Real-Money Returns and Benchmark Analysis

> Discover AI Berkshire's verified performance track records. See real-money returns of +69.29% (2024) and +66.38% (2025), outperforming the S&P 500 significantly. Analyze audited results.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: performance
- Published: 2026-07-05

---

**AI Berkshire has generated verified annual returns of +69.29% (2024) and +66.38% (2025) in a real-money, audited portfolio, outperforming the S&P 500 by over 46 percentage points each year according to the repository’s documented track record.**

The documented performance track records of AI Berkshire represent one of the few publicly auditable AI-driven investment frameworks backed by live brokerage data. Unlike backtested simulations, these figures derive from a real-money portfolio hosted on Futu Securities, with cumulative live returns exceeding ¥1.46 million over the two-year period. All verification screenshots and calculation methodologies are stored directly in the `xbtlin/ai-berkshire` repository.

## Verified Annual Returns vs. Global Benchmarks

The framework’s performance is benchmarked against major global indices to demonstrate absolute and relative alpha generation.

| Year | Full-Year Return | Benchmark Comparison |
|------|------------------|----------------------|
| **2024** | **+69.29%** | Hang Seng +17.67% · S&P 500 +23.31% · CSI 300 +14.68% · NASDAQ +28.64% |
| **2025** | **+66.38%** | Hang Seng +27.77% · S&P 500 +16.39% · CSI 300 +17.66% · NASDAQ +20.36% |

In **2024**, the AI Berkshire portfolio beat the S&P 500 by **46 percentage points**. In **2025**, the outperformance margin widened to **50 percentage points**. These figures are documented in the **Real Track Record** section of [`README_EN.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README_EN.md) and supported by live account screenshots.

## How the Track Record Is Generated and Audited

The performance numbers rely on a rigorous validation pipeline that prevents calculation errors and data inconsistencies.

### Real-Money Portfolio Verification

The repository stores actual brokerage screenshots that serve as audit trails:

- `assets/2024-returns.jpg` – Screenshot of the 2024 live portfolio performance
- `assets/2025-returns.jpg` – Screenshot of the 2025 live portfolio performance

These images capture the Futu Securities account interface, confirming the documented performance track records reflect real market executions rather than theoretical simulations.

### Exact-Decimal Validation Engine

All financial calculations use [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py), which enforces exact-decimal arithmetic to eliminate floating-point errors. This engine cross-validates market-cap calculations, position sizing, and return aggregations before they are published in the track record.

## Reproducing the Performance Methodology

You can invoke the same skills and tools that generate the research powering these verified returns.

### Running the Investment Research Pipeline

The `investment-research` skill executes multi-agent analysis that feeds into portfolio decisions tracked in the performance record.

```bash

# Install the Codex skills (once)

git clone https://github.com/xbtlin/ai-berkshire.git
cd ai-berkshire
./scripts/install-codex-skills.sh

# Launch analysis on a target company (e.g., Tencent)

codex use investment-research "Tencent"

```

This skill pulls live data and validates financial numbers using [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py), outputting research memos that inform the portfolio’s constituents.

### Calculating Live Portfolio Returns

The `portfolio-review` skill aggregates holdings and computes real-time performance using the same engine that produces the audited track record.

```bash
codex use portfolio-review \
  "Tencent, 30%" "Meituan, 20%" "Kweichow Moutai, 20%" "Cash, 30%"

```

This command fetches latest market prices via the built-in web scraper and calculates returns using exact-decimal methods.

### Cross-Validating Financial Calculations

Verify individual calculations using the validation engine directly:

```bash
python3 tools/financial_rigor.py verify-market-cap \
  --price 510 --shares 9.11e9 --reported 4.65e12 --currency HKD

```

This reproduces the rigorous cross-validation steps underlying the track record’s credibility.

## Key Source Files Behind the Track Record

- **[`README_EN.md`](https://github.com/xbtlin/ai-berkshire/blob/main/README_EN.md)** (section *Real Track Record*) – Central documentation for audited returns and benchmark tables
- **`assets/2024-returns.jpg`** – Screenshot evidence of 2024 live portfolio performance
- **`assets/2025-returns.jpg`** – Screenshot evidence of 2025 live portfolio performance  
- **[`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py)** – Exact-decimal calculation engine ensuring figure reliability
- **[`skills/portfolio-review.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/portfolio-review.md)** – Skill that aggregates holdings and computes portfolio-level returns
- **[`skills/investment-research.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/investment-research.md)** – Core research skill generating the deep-analysis reports driving investment decisions

## Summary

- AI Berkshire’s documented performance track records show **+69.29%** (2024) and **+66.38%** (2025) returns in a real-money portfolio.
- The framework outperformed the S&P 500 by **46–50 percentage points** annually against major global indices.
- All claims are supported by live brokerage screenshots stored in `assets/2024-returns.jpg` and `assets/2025-returns.jpg`.
- Exact-decimal validation via [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) guarantees calculation accuracy.
- Users can reproduce the methodology using `investment-research` and `portfolio-review` skills.

## Frequently Asked Questions

### Is AI Berkshire's performance track record independently audited?

The track record is documented through real-money brokerage screenshots from Futu Securities rather than third-party audit statements. The repository stores these screenshots in `assets/2024-returns.jpg` and `assets/2025-returns.jpg`, providing visual verification of the live account performance. All calculation methodologies are open-source and stored in [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py) for public inspection.

### How does AI Berkshire ensure the accuracy of its return calculations?

The framework uses [`tools/financial_rigor.py`](https://github.com/xbtlin/ai-berkshire/blob/main/tools/financial_rigor.py), an exact-decimal calculation engine that eliminates floating-point errors common in financial software. This tool cross-validates market-cap figures, position weights, and aggregated returns before publication. The `portfolio-review` skill invokes this engine when computing real-time portfolio performance.

### Can investors reproduce the decisions that generated these returns?

Yes. The repository provides Codex skills that replicate the research process. Users can run `codex use investment-research` to generate deep-analysis reports and `codex use portfolio-review` to calculate portfolio-level returns using the same validation logic. This allows independent verification of the methodology behind the documented performance track records.

### What benchmarks does AI Berkshire use for performance comparison?

The track record compares returns against four major global indices: the **Hang Seng Index**, **S&P 500**, **CSI 300**, and **NASDAQ**. The framework has outperformed all four benchmarks in both 2024 and 2025, with particular outperformance versus the S&P 500 exceeding 46 percentage points each year.