NemoClaw Knowledge Wiki

Chat Deep Research update. Channel Rob the AI Guy

3 min read

Chat Deep Research update. Channel Rob the AI Guy

https://www.youtube.com/watch?v=XmKD8BGiC8w Here is a Markdown summary of the video regarding the new ChatGPT Deep Research update.

ChatGPT Deep Research Update: Features & Walkthrough

This update provides users with granular control over research sources and integrates directly with third-party applications, making it significantly more powerful than the legacy Deep Research model.

🚀 How to Access

  • Open ChatGPT.
  • Locate Deep Research on the left-hand sidebar (grouped with Pulse, Images, and Apps).
  • Alternatively, click the ”+” (plus) icon in the chat bar to toggle Deep Research on.

🛠 Key New Features

1. App Integration

You can now grant ChatGPT Deep Research access to external apps to query their specific databases rather than just “searching the web.”

  • Supported Apps: Expedia, Slack, Zillow, Google Drive, GitHub, HubSpot, Stripe, and over 60 others.
  • Categories: The apps are sorted into Featured, Lifestyle, and Productivity.
  • Functionality: This allows the AI to generate reports based on your private data (e.g., Amplitude, Conductor) or specific platform data.

2. Granular Site Control

You have total control over where the AI looks for information.

  • Specific Sites: You can instruct it to search only specific URLs.
  • Manage Sites: You can exclude specific sites or domains.
  • List Management: You can copy, save, or clear lists of sites for future use.

3. “Human-in-the-Loop” Workflow

Unlike the legacy version, this process is interactive.

  • Transparency: You can see the AI’s steps (Collecting, Extracting, Identifying, Comparing, Synthesizing).
  • Intervention: You can Pause or click Update during the research process to:
    • Ask follow-up questions.
    • Make adjustments.
    • Add new constraints or advice mid-research.

4. Comprehensive Reporting

The output is a long-form, academic-style report (can take 20–45 minutes to generate).

  • Content: Includes executive summaries, citations, data tables, and “Mermaid” causal flowcharts.
  • Export Options: Reports can be exported to PDF or Word.
  • Depth: In the examples shown, the AI performed over 400 searches and used 20+ citations for a single report.

💡 Use Case Examples

Goal: Analyze flight and hotel price fluctuations from NYC (JFK) to Florida (PBI) for a 5-star trip throughout the year.

  • Process: The AI queried the Expedia database directly.
  • Result: It produced a report showing:
    • Seasonal price dynamics.
    • Lowest and highest combined monthly costs.
    • Peak-to-trough price swings.
    • Data tables for every month of the year.

Use Case 2: Real Estate Relocation (Zillow Integration)

Goal: Determine the best towns in Montana to move to with specific criteria (1M budget, >3,000 sq ft, >5 acres of land).

  • Process: The AI accessed Zillow to find live inventory data rather than generic articles.
  • Result: It generated an executive summary including:
    • Specific towns with the highest inventory (e.g., Kalispell, Helena).
    • Median list prices and square footage per town.
    • Specific example listings with links.
    • Assumption and limitation notes regarding the data.

⚠️ Important Note

This new version of Deep Research is slower than the legacy version (taking 20+ minutes per report) because it mimics a human research workflow. However, the depth of data and the ability to customize sources make the output significantly more usable.

Graph View