Stock Research Automation

The Backstory:

A few of us based in the Philippines have been trading US stocks as a hobby for a while. US markets open at 9:30 PM our time, which means we have our evenings free to prepare before the action starts.

Every night before market open, we'd manually:

• Scan through market losers and gainers from the previous session
• Read earnings calendars and analyst reports
• Check Reddit, Twitter, and financial newsletters
• Pull up charts on Yahoo Finance
• Cross-reference news from multiple countries
• Write up our analysis in a shared spreadsheet

The goal was always the same: figure out what stocks have potential for short-term gains (weeks to months) and be ready to act in the first few hours of market open.

It worked, but it was time-consuming. We'd spend 2-3 hours each evening just gathering data before we could even start analyzing. So we thought: why not automate the boring parts?

This case study walks through what we built — a fully autonomous research system that runs before US market open and delivers ranked stock picks just in time for trading.

What We Built:

The Basic Idea

Two scheduled workflows that run back-to-back before US market opens:

1. Phase 1 (8PM PHT): Gather Intelligence from 8 Different Angles
2. Phase 2 (9PM PHT): Run technical analysis and rank everything

By 9:30 PM when US markets open, we have a prioritized list of short-term stock ideas with entry points, stop losses, and rationale — ready to act on.

Architecture

The Research Streams

We basically replicated our manual research process, but with AI agents handling each piece:

What We Used To Do	AI Agent	Data Sources
Check market losers for rebounds	Losers Analyst	Yahoo Finance losers, news
Scout pre-earnings plays	Earnings Scout	Earnings calendars, estimates
Track momentum stocks	Momentum Tracker	Gainers data, volume
Read international news	Global News Analyst	Reuters, Bloomberg, regional media
Check local news (JP, US, CN, DE, IN)	Local Intelligence	Country-specific sources
Browse Reddit/Twitter/TikTok	Social Sentiment	Reddit, Twitter/X, Weibo, TikTok
Track consumer spending trends	Macro Analyst	Economic data, spending patterns
Read financial newsletters	Newsletter Analyst	Curated newsletter content
Do chart analysis	Technical Strategist	Yahoo Finance (9 endpoints)

Plus three Ranking Agents that consolidate everything into a Top 10 list, and an Editor Agent that formats the final output.

Tech Stack

Component	What it does
n8n	Orchestrates everything — scheduling, flow control, error handling
GPT-5.1	Powers the research and analysis agents
Google Gemini	Handles final editing (cheaper for simple formatting)
Serper MCP	Real-time Google search and webpage scraping
Yahoo Finance MCP	Stock prices, financials, analyst ratings, options data
Google Sheets	Stores intermediate data and final picks
Todoist	Creates a task each evening with the day's picks

How It Actually Works

Phase 1: Intelligence Gathering

Each research stream runs through the same pattern:

1. Deep Search Subworkflow — An AI agent with access to Google Search and webpage scraping. We give it a specific prompt (e.g., "find top 10 market losers with rebound potential") and it goes hunting.
2. Smart Summarization — If the output is too long (>45K characters), another agent condenses it while preserving all the important numbers and sources.
3. Store to Sheets — Results go into Google Sheets with a processed=false flag.

Phase 2: Analysis & Ranking

1. Read unprocessed records from Phase 1
2. Technical Analysis — Each stream goes through a "Meta-Analyst" agent that:
   • Pulls Yahoo Finance data (prices, financials, recommendations)
   • Calculates technical indicators (RSI, MACD, moving averages)
   • Merges with the fundamental analysis from Phase 1
3. Ranking — Three GPT-5.1 agents rank candidates and pick the Top 10
4. Editing — Google Gemini formats everything nicely
5. Output — Final picks saved to Sheets + Todoist task created

Outputs

Sample Output

Here's what a typical recommendation looks like internally between the agents:

            🔢 Rank: 1
            📈 BABA – Alibaba Group (China, E-commerce/Cloud)
            📊 Sector: Consumer Discretionary / Technology
            ⏳ Target Horizon: Short-Term (Pre-Earnings)
            ✅ Adjusted Confidence: 78%
            🎯 Risk-Reward Ratio: 3.2:1
            🧮 Composite Score: 8.4/10

            🧠 Combined Rationale:
            Alibaba presents a compelling pre-earnings setup ahead of Nov 25 results.
            Strong Buy consensus with $180-230 price targets (25-40% upside). Active
            buyback program with $4B deployed YTD. 72% YTD rally showing institutional
            re-rating in progress.

            Technically: confirmed uptrend, price above 50/200 MAs. RSI at 58 (momentum
            building, not overbought). MACD bullish and expanding.

            📌 Key Technical Dashboard:
            | Indicator        | Value              | Implication           |
            |------------------|--------------------|-----------------------|
            | Market Structure | Higher Highs/Lows  | Confirmed Uptrend     |
            | 50/200 MA        | Price > Both       | Bullish Alignment     |
            | RSI (14)         | 58, Rising         | Momentum Building     |
            | MACD             | Bullish, Expanding | Trend Acceleration    |
            | Analyst Consensus| 85% Buy            | Strong Street Support |

            🚦 Action Plan:
            - Entry: Break above $115 with volume
            - Stop-Loss: $108 (below 20-day MA)
            - Target 1: $120 | Target 2: $135
            - R/R to T1: 3.2:1

            🛑 Risks:
            - China regulatory headlines
            - Cloud growth could disappoint
            - Broad market risk-off
          

Actual Output Posted to Twitter

Here are the actual stock picks posted on our Twitter: @siliconesignals

Silicone Signals @siliconesignals · 20h
Automated by @dynameyes
Breaking It Down: The Top 10 Picks in Plain English, December 08, 2025

Rank 1: $NVDA (NVIDIA Corp.)
The undisputed king of the "AI arms race," building the advanced chips that power artificial intelligence worldwide. Revenue up >60%, and they're more than just a chip maker — they're a full AI system provider. The stock is in a strong uptrend; analysts suggest waiting for a small dip to buy rather than chasing high prices.
Watch out for: If AI spending suddenly slows down or new export rules limit sales, the stock could drop quickly.

Rank 2: $MU (Micron Technology Inc.)
While NVIDIA provides the "brains," Micron provides the memory that AI computers desperately need. Prices for memory chips are rising due to shortages, which is great for Micron's profits. A smart way to bet on the AI boom with a stock that's still reasonably priced compared to its growth.
Watch out for: This industry is famous for "boom and bust" cycles; if supply catches up too fast, prices could fall.

Rank 3: $MDLZ (Mondelez International, Inc.)
The maker of Oreos and Cadbury is looking tasty again because the cost of ingredients like cocoa and coffee has dropped significantly. A "defensive" pick offering a steady dividend and a chance to buy while the price is relatively low.
Watch out for: Grocery stores might demand lower prices, which could eat into those extra profits.

Rank 4: $TSM (Taiwan Semiconductor Manufacturing Co.)
The world's most important chip factory — they actually build the chips for companies like NVIDIA and Apple. Because they're the bottleneck for all advanced AI technology, they have huge pricing power. A "must-own" infrastructure play.
Watch out for: Geopolitical tension between Taiwan and China is the biggest risk here, regardless of how well the business is doing.

Rank 5: $DAL (Delta Air Lines Inc.)
Travel is booming, with planes full and ticket prices healthy, yet Delta's stock price is still considered "cheap" relative to its profits. A solid pick for the short-term travel season.
Watch out for: Sudden spikes in oil prices or a recession could hurt travel demand and profits very quickly.

Rank 6: $AMZN (Amazon.com, Inc.)
Amazon is firing on all cylinders: holiday shopping is breaking records, and their Cloud business (AWS) is growing fast. Experts are particularly excited about Amazon building its own custom AI chips, making it less dependent on others.
Watch out for: If consumer spending dries up after the holidays, their retail profits could take a hit.

Rank 7: $TJX (The TJX Companies — TJ Maxx/Marshalls)
When regular stores have too much inventory, they sell it to TJ Maxx, who sells it to bargain hunters. This business model works great when shoppers are looking to save money. Trading near all-time highs.
Watch out for: If the economy crashes too hard, even discount shoppers might pull back on spending.

Rank 8: $LION (Lionsgate Studios Corp.)
Home to franchises like John Wick and The Hunger Games, this studio is a potential takeover target. Experts think Lionsgate is undervalued and could jump in price if a deal or partnership is announced. A riskier, "event-driven" trade.
Watch out for: If no deal happens, the stock could struggle, as their regular financials aren't as strong as the big players.

Rank 9: $RELIANCE.NS (Reliance Industries Ltd.)
A massive Indian conglomerate with hands in everything from energy to retail and telecom. A way to bet on India's fast-growing economy (GDP ~8%) and the boom in digital spending.
Watch out for: Government regulation changes in India or global shifts away from emerging markets could hurt the stock.

Rank 10: $MELI (MercadoLibre, Inc.)
Often called the "Amazon of Latin America," this company dominates e-commerce and payments in the region. The stock recently dipped, offering a chance to buy a high-growth company at a discount.
Watch out for: Doing business in Latin America involves currency risks and volatile economies, which can make the stock price jumpy.

What We Learned

Things That Worked Well

• Prompt engineering matters a lot — The quality of output depends heavily on how you frame the agent's role and constraints
• Using multiple LLMs — GPT-5.1 for heavy reasoning, Gemini for simple formatting saved costs
• Smart summarization — Keeping outputs under token limits without losing information
• Rate limiting in prompts — Telling the agent "max 10 API calls" actually works

Things We'd Do Differently

• The Vars node — We built infrastructure for adaptive load management but never actually wired it up. Classic over-engineering.
• Better error handling — Sometimes an agent times out and we lose that stream's data for the day

Numbers

Metric	Before (Manual)	After (Automated)
Time spent gathering data	2-3 hours/day	0
Stocks analyzed	~20-30	100+
Data sources per pick	3-5	15-25
Consistency	Variable (depends on mood)	Same process every day

Skills & Tech Used

AI/LLM

• Multi-agent architecture
• Prompt engineering
• Output parsing

Workflow Automation

• N8N orchestration
• Scheduled execution
• Error handling

Financial Data

• Yahoo Finance API integration
• Technical indicator analysis
• Fundamental analysis

Integration

• MCP (Model Context Protocol)
• Google Sheets API
• Todoist API
• SSE (Server-Sent Events)

Wrap Up

What started as "let's automate our nightly research routine" turned into a pretty capable system. It's not perfect — sometimes the AI hallucinates or misses obvious things — but it's a solid starting point that saves us hours every evening.

The timing works out nicely for us in the Philippines. The system runs at 8-9 PM while we're winding down from dinner, and by 9:30 PM when US markets open, we have a shortlist ready. Instead of scrambling to research during the first hour of trading, we can focus on execution and monitoring our picks.

For short-term trades spanning weeks to months, having this kind of systematic daily analysis helps us catch momentum early and avoid chasing moves we're late to.