The Killer App

The "Killer App" for SFRS isn't just a better validator—it's Longitudinal AI Analysis.

Because SFRS is token-optimized and JSON-native, you can perform deep historical analysis that is technically impossible or economically prohibitive with legacy XBRL.

The Problem: The "Token Tax"

Financial analysts increasingly use Large Language Models (LLMs) to scan filings. However, XBRL (XML) carries a massive "token tax":

• Verbose Tags: <us-gaap:NetIncomeLoss contextRef="ctx_2023" unitRef="USD" decimals="-6"> is 42 tokens.
• JSON Equivalent: "netIncome": 3400 is 4 tokens.

When analyzing a single company's 10-year history, this difference is the difference between fitting in a context window and failing.

Comparison: 10 Years of 10-Ks

Try the Demo

We've provided a script to synthesize a 10-year longitudinal SFRS record and estimate the token savings.

node scripts/killer-app-demo.mjs

Ticker Explorer: Real-World Data

You can also explore real-world company data by ticker. The Ticker Explorer maps a public ticker to its SEC CIK, fetches recent filing metadata, and demonstrates how that data is condensed into SFRS.

node scripts/ticker-explorer.mjs NVDA

This script will:

1. Map NVDA to CIK 0001045810.
2. Fetch the metadata for the last five 10-K filings.
3. Generate a sample longitudinal SFRS file at dev/active/nvda-history.sfrs.json.

Real-World AI Analysis via API

You can programmatically build longitudinal datasets for any public company using the SFRS API and the SEC'gar archive.

1. Identify Filing Archive

Map a ticker to its SEC CIK and locate the .xml XBRL instance for the desired periods.

# Example for NVDA 2024 10-K
export XBRL_URL="https://www.sec.gov/Archives/edgar/data/1045810/000104581024000029/nvda-20240128_htm.xml"

2. Convert to SFRS via API

POST the XBRL URL to the SFRS conversion endpoint to get a dense JSON payload.

curl -X POST https://api.sureshake.org/api/v1/convert/from-xbrl \
  -H "Content-Type: application/json" \
  -d '{
    "sourceUrl": "'$XBRL_URL'",
    "options": { "taxonomy": "us-gaap-2024" }
  }' > nvda-2024.sfrs.json

3. Aggregate and Analyze

Repeat for 5 years and combine the payloads. The resulting file will be ~98% smaller than the original XBRL, allowing you to send the entire 5-year history to an LLM in a single request:

// Sample Node.js script using Google Generative AI
import { GoogleGenerativeAI } from "@google/generative-ai";

const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);
const model = genAI.getGenerativeModel({ model: "gemini-3.1-pro-preview" });

const history = [/* Combined SFRS Payloads from 5 years of filings */];

const prompt = `You are a financial analyst specializing in SFRS data. 
Analyze the following 5-year longitudinal financial history for trends in 
profitability and asset efficiency: ${JSON.stringify(history)}`;

const result = await model.generateContent(prompt);
console.log(result.response.text());

Sample Analyst Prompt

Once you have the longitudinal SFRS data, you can feed it to an LLM with a prompt like this:

"I am providing 10 years of financial data for Acme Corporation in SFRS format. Please identify the trend in R&D intensity relative to revenue and predict the 2025 net income based on the 8% historical CAGR."

Why This Matters

1. Benchmarking: Compare a company against 50 peers in a single context window.
2. Anomaly Detection: Find "drift" in accounting policies across a decade in seconds.
3. Real-time Valuation: Feed dense financial data into an agentic workflow for instant valuation modeling.

SFRS makes financial data machine-useful, not just machine-readable.