Think about Agent mode as having your own intern or analyst.
I'm going to show you this week how I used ChatGPT Agent mode to scrape for investment ideas:
For this example, I wanted to monitor trade ideas from the last 6 months from X user Scott Grossman @srg444 - he is an ex-Magentar PM and criminally underfollowed account.
You are my copytrading analyst.
Your job is to analyze the last 6 months of tweets (including replies) from the account @srg444 (https://x.com/srg444/with_replies).
Extract all relevant tweets where a stock or company is mentioned, and create a structured table with the following:
Output Table Columns:
Company / Ticker Sentiment:
Bullish, Bearish, Neutral, or Unclear Summary of Tweet (1–2 sentences)
Date of Tweet
Link to Tweet
Rationale or Signals (any hints, thesis, language used that signal conviction or actionable ideas)
Rules:
Include both original tweets and replies.
Only include tweets that mention specific companies, tickers, or clearly imply stock positions.
Use language cues (e.g., “loading,” “shorting,” “looks cheap,” “betting on,” “AI winner”) to determine sentiment.
If sentiment is not clearly stated, mark as “Unclear.”
Format:
Deliver results in a well-formatted Excel file
Bold headers
Colored header row (dark navy background with white text)
Use alternating row shading
Ensure links are clickable
Focus only on tweets related to public equities or clear trade ideas.
Overall I think it did an ok job - there are names ChatGPT Agent missed but I expect the consistency to improve over time (they make upgrades based on user data / feedback). Keep in mind that the web search function was not great at roll-out but now rarely makes errors.