Demo: AI Agents in Market Analysis

Hey, it's Matt. Welcome to AI Street Markets, where I break down AI Investing tools.

Was this email forwarded to you? Sign up below:

In This Edition:

• AI Demo: ScalarField.io 

  • Get 👉️ Complimentary 👈️ Access: 3-month promo code below

• Markets Wrap & Prompt: S&P hits new highs, Netflix jumps, oil slides

AI AGENTS VS RAG

AI has historically been deterministic and fixed.

But ChatGPT is probabilistic and flexible.

The first gives consistent but rigid outputs; the second offers creative, but unpredictable results.

Businesses want the best of both: reliable answers but the ease of use of ChatGPT.

The question is: how do you combine the two?

Right now, one of the current solutions is retrieval augmented generation (RAG), which combines traditional document retrieval with language models. Here's how it works:

1. Document Integration: Companies integrate their existing documentation into the system

2. Contextual Processing: The AI analyzes queries against this verified information

3. Enhanced Response: The system generates answers grounded in authenticated data

Key Advantage: RAG provides reliable, fact-based responses while maintaining natural language capabilities.

Primary Challenge: Even with accurate source material, AI may still misinterpret complex data relationships.

LLM -> RAG System -> Data -> LLM 

AI Agents

Agent-based systems take a different approach by breaking complex tasks into smaller, specialized components. This architecture involves:

1. Task Distribution: Different AI agents handle specific aspects of the query

2. Specialized Processing: Each agent focuses on its area of expertise (data retrieval, calculation, etc.)

3. Coordinated Response: Results are combined into a coherent answer

Key Advantage: Higher accuracy through specialized processing and structured workflows.

Primary Challenge: Increased system complexity and computational requirements.

LLM -> Specialized Agents -> Structured Data Sources -> LLM 

Demo: ScalarField's Agent-Based Analysis Platform

One platform trying the agent approach is ScalarField.io. I got a quick demo this week by the company’s founder, Amandeep Singh, PhD, who has a history of building trading technology. He traded interest rates at Goldman Sachs. Before that, he was a quant trader at Tower Research. Singh is also a professor at the University of Washington.

In the demo video above, which I’ve sped up 3x, you’ll see how ScalarField uses agents for different tasks, guides my prompt for more specificity and suggests follow up analysis.

Here’s my open-ended prompt:

Which small cap companies are attracting a growing number of institutional buyers and what could be sparking their interest?

ScalarField asks me for more specificity:

And you can see how it’s processing the request:

With traditional tools this analysis would take at least a few hours. In this demo, ScalarField took about five minutes. Here are the key results:

• PBF Energy led in net institutional buying

• Most top companies showed consistent buying across 7-11 months

• Energy sector dominated institutional interest

• Buy ratios above 90% suggest strong institutional conviction

And here’s the link to the full analysis and prompts.

One advantage I think of this approach is that the data results are quantifiable since you can access the code used by ScalarField if you have a pro account.

MARKETS WRAP  

The following was generated by Perplexity. I’ve shared the prompt below so you can try it yourself.

Market Roundup: January 19-26, 2025

Top Market Moves

• S&P 500 Reaches New High: The S&P 500 closed at a record high of over 6100, marking a weekly gain of 2.04%. This surge followed President Trump's call for interest rate cuts and lower oil prices, contributing to positive investor sentiment across the market.

• Dow Jones Industrial Average Gains 2.48%: The Dow rose by 400 points during the week, reflecting strong performance driven by optimism surrounding economic growth and corporate earnings.

• Nikkei 225 Declines Post Rate Hike: Following the Bank of Japan's decision to raise interest rates by 25 basis points to 0.5%, the Nikkei 225 index experienced a slight decline, although it remains up 3.9% for the week due to earlier gains.

• Oil Prices Drop: Crude oil prices fell by approximately 4% to $74.29 per barrel amid concerns over demand and geopolitical tensions, impacting energy sector stocks negatively.

• Bitcoin Climbs 2.8%: Bitcoin rose to $104,046, reflecting growing interest and investment in cryptocurrencies despite market volatility.

Key Economic Updates

• US Jobless Claims Increase: Initial jobless claims rose by 6,000 to 223,000, indicating potential labor market weaknesses but remaining near historic lows.

• UK Unemployment Rate Rises: The unemployment rate in the UK increased to 4.4% in November, its highest level in three months, raising concerns about economic stability in the region.

• China's Economic Stabilization Efforts: The People's Bank of China maintained its lending rates as part of ongoing efforts to stabilize the economy amid deflationary pressures and weak consumer confidence.

• ECB Expected to Cut Rates: The European Central Bank is anticipated to lower its deposit rate from 3% to 2.75% at its upcoming meeting as part of measures to support sluggish economic growth.

• Japanese Corporate Earnings Strong: Major Japanese corporations reported robust earnings for Q3, with Disco Corp. showing a remarkable operating profit increase of 54.6% year-on-year, signaling strong performance amid market fluctuations.

Corporate Headlines

• Netflix Surges After Earnings Report: Netflix's shares jumped by 16% following a strong quarterly earnings report that revealed a rise in global subscribers to 302 million and a revenue increase of 16% year-on-year.

• GE Aerospace Exceeds Revenue Expectations: GE Aerospace reported revenues of $7.65 billion from commercial engines and services, surpassing analyst expectations and projecting low-double-digit revenue growth for 2025.

• Tesla Stock Declines: Tesla's shares fell by 3.7% after reports indicated a decline in iPhone shipments in China, reflecting broader concerns about supply chain issues and market demand.

• Cybozu Announces Increased Dividends: Cybozu surprised investors with an announcement of increased dividends from 14 yen to 30 yen per share, reflecting improved profit forecasts and confidence in future performance.

MARKETS PROMPT

Copy and paste the below prompt into Perplexity (it’s free). Make sure to select Pro here:

Task: 

Create a concise, data-driven market roundup covering the most critical developments from the last 7 days. Use New York EST as your time zone. Format: "Market Roundup: [DATE RANGE]"

INTERNAL REQUIREMENTS (do not display):

1. Source Diversity: Include at least 8–10 credible sources, prioritizing:

   • Global: Bloomberg, Reuters, WSJ, FT, CNBC, Yahoo Finance, Nikkei Asia, South China Morning Post, Economist

   • Regional: Economic Times (India), Business Standard (India), CNBC TV18 (India), AFR (Australia), Handelsblatt (Germany)

   • Primary Sources: Fed/ECB statements, SEC filings, earnings reports, government press releases

   • Avoid: Unverified blogs, social media, or low-authority outlets.

2. Precision:

   • Specific numbers: Include percentages, price levels, volumes, or policy rates.

   • Global scope: Highlight developments from major markets (US, EU, Asia) and key emerging economies.

3. Sections & Structure:

   • Top Market Moves (3–5 points): Focus on indexes, sectors, commodities, or currencies with >1% daily moves or record highs/lows.

   • Key Economic Updates (3–5 points): Prioritize GDP, inflation, employment data, central bank decisions, or geopolitical policies.

   • Corporate Headlines (3–5 points): Major earnings surprises (>5% deviation), M&A deals (>$1B), or CEO/board changes.

4. Formatting Rules:

   • Source tags: Use uppercase abbreviations (e.g., REUTERS, SEC, WSJ).

   • Consistency: Begin each bullet with bolded key metric/event, followed by concise context.

   • No markdown: Use plain text with clear section headers

Hope you found this edition helpful. As always, if you have any questions, reply to this email.

See you next week!