NemoClaw Knowledge Wiki

AI for deep research - a comparison

4 min read

AI for deep research - a comparison

https://www.youtube.com/watch?v=meg8PU7qnz4 The video, presented by Robert Benjamin, compares various AI tools for “deep research” and then outlines several “insane use cases” for these capabilities, focusing on the DeepAgent tool available through Abacus.AI. AI Tool Comparison for Deep Research: Robert tested six AI tools with the prompt: “Please craft a research report about how global warming is affecting certain animal populations in the United States. Please include plenty of graphs and photos that support what you find.”

  • Methodology:
    • Some tools (Gemini, ChatGPT, DeepAgent) asked clarifying questions before proceeding, which Robert found beneficial for tailoring the research.
    • Others (Grok, Genspark, Manus, Perplexity) immediately started the research without further input.
  • Results:
    • Gemini: Researched 169 websites but ultimately failed to fulfill the request, stating, “I’m having a hard time fulfilling your request.” Robert found this “incredibly disappointing.”
    • ChatGPT: Took 7 minutes, 33 sources, and 90 searches. It produced a text-heavy report with very few and small images/graphs, not fully meeting the visual requirements. Robert was “disappointed.”
    • Grok: Did not provide any graphs or photos, only a small table, despite the explicit request in the prompt. Robert was “even more disappointed.”
    • Genspark: Created a “PowerPoint” style output with graphs and photos, but lacked substantial text, resembling a website or presentation more than a research report.
    • Manus: Generated a lengthy 44-page PDF report. However, it completely missed the requirement for graphs and photos. Robert felt this would be “screwed” for real-world application due to the missing visuals.
    • Perplexity: Produced a “pretty short” report (48 sources) with some graphs and photos. It was decent but lacked depth.
    • DeepAgent (Winner): This tool “incredibly impressed” Robert. It produced a comprehensive 23-page PDF report. Crucially, it successfully included a wealth of relevant graphs (e.g., temperature changes, habitat loss, biodiversity maps) and descriptive photos (e.g., drought, coral bleaching, polar bears), directly fulfilling the prompt’s visual requirements.

Robert’s Ranking:

  1. DeepAgent
  2. Perplexity
  3. Manus (due to report length, despite missing visuals)

Five Insane Use Cases for DeepAgent’s Research Capabilities: Robert highlights DeepAgent’s strength in combining deep research with other AI agent features for powerful applications:

  1. Market Research:
    • Prompt Example: “Do an in-depth research about market growth and future opportunities in the fitness and nutrition industry globally. Use bullets, tables and charts to make the report clearer.”
    • DeepAgent Output: A comprehensive PDF report with market sizing, growth projections, regional analysis, investment opportunities, and relevant charts and tables.
    • Use Case: Create detailed market research reports for personal interest, business strategy, or even as lead magnets to attract clients by providing valuable industry insights.
  2. Travel Itinerary Planning:
    • Prompt Example: “I am planning a budget solo backpacking trip to Tokyo. Please create a 5 day itinerary.”
    • DeepAgent Output: Interactively asks clarifying questions (budget range, interests, time of year, accommodation preference, specific areas). Then generates a detailed PDF itinerary covering best neighborhoods, budget accommodation recommendations, transportation costs, daily activities, food guides, safety tips, and includes maps and images.
    • Use Case: Generate personalized and detailed travel plans for any destination, complete with practical information and visuals, saving immense time on manual research.
  3. In-depth Competitive/Industry Analysis:
    • Prompt Example: “Do an in-depth research around EV battery manufacturers globally, and provide a competitive study around them. Cite source of information for all the information you use. The report should have one section to rate the manufacturers across various relevant parameters. Provide the report in pdf format.”
    • DeepAgent Output: Asks clarifying questions (focus on global/local players, specific battery technologies, key comparison parameters, future outlook, geographic regions). Produces a 53-page PDF report detailing market share, production capacity, technology, R&D, partnerships, and manufacturing facilities for major global EV battery manufacturers, including visual charts and cited sources.
    • Use Case: Conduct highly detailed analyses for academic pursuits (PhD, Master’s), professional work, or gaining expertise in complex industries, providing comprehensive overviews faster than traditional methods.
  4. Social Media Trend Reports:
    • Prompt Example: “Generate a report summarizing the top trending topics on social media and across the internet over the past 1 week. For each trending topic, the agenda of this report is know the topics to post about on my socials to increase engagement. please include the following: Topic Name…, Platform(s)…, General Sentiment…, Quanlitative Metrics…, Engagement levels…, Regional/geographic popularity…, Summary of Discussion…”
    • DeepAgent Output: A detailed report categorizing trending topics, identifying platforms, sentiment analysis, quantitative metrics (posts/mentions, engagement levels, growth rate), geographic distribution, and a discussion summary.
    • Use Case: Inform social media content strategy, identify emerging trends, and understand audience sentiment for specific niches to increase engagement and relevance.

Robert concludes by recommending DeepAgent (available through Abacus.AI starting at $10/month) for its superior deep research capabilities, noting its ability to interact with various LLMs and build custom AI agents.

Graph View