The Critical Insight & Gap-Finder

• 

  This image visualizes the core problem (cluttered research) and your solution (a concise report).

• 

  This image focuses on Comet Browser as the active agent, performing the intelligent workflow.

• 

  This image highlights the most valuable part of your project: the generated "Research Gaps" as a thesis.

Inspiration

As a student, my biggest bottleneck isn't finding information—it's synthesizing it. I've spent countless nights with 20+ tabs open, drowning in research articles, but still having no idea what my actual argument should be. I was a human "copy-paste" machine, collecting facts but struggling to find the critical insight.

I was inspired to build The Critical Insight & Gap-Finder to automate the one thing I couldn't: critical thinking. I wanted a tool that wouldn't just summarize what others said, but would actively find the debates, the counter-arguments, and the "Aha!" gaps between the sources—the very things that form a strong thesis.

What it does

The Critical Insight & Gap-Finder is a single, powerful automation workflow for the Comet Browser. A user provides a single research topic (e.g., "AI in high-school essays"), and the Comet AI agent executes a complex, multi-step workflow:

1. It automatically searches for and reads the top-cited core research on the topic.
2. It then performs a "Criticism Search" to find the strongest counter-arguments.
3. Finally, it opens a new Google Doc and generates a structured "Critical Insight Report" that includes:
   • Key Themes & Consensus
   • Major Debates & Criticisms
   • A list of "Potential Research Gaps" that serves as a perfect starting point for a thesis.

It turns 10+ hours of research into a 90-second automated process.

How I built it

With the hackathon's tight deadline, my "code" isn't Python; it's a single, complex Master Prompt that I iterated on and refined. My process had three phases:

1. V1 (The Simple Summarizer): My first attempt was a basic prompt: "Find 5 articles on a topic and summarize them." This worked, but it wasn't innovative—it just automated the "copy-paste" part.
2. V2 (The Pivot): This was the breakthrough. I realized the key to a good paper isn't the consensus, it's the conflict. I updated my workflow to have the Comet agent perform a "Criticism Search"—a dedicated step to find the strongest counter-arguments.
3. V3 (The Final "Master Prompt"): I finalized the workflow by asking the agent to synthesize both the core research and the criticisms. The final output isn't a list; it's the structured "Critical Insight Report" that delivers true value.

Challenges I ran into

The main challenge was "prompt engineering." Agentic AIs are incredibly powerful but can be brittle. An instruction like "read this page" is too vague. I had to learn to be precise, giving the agent a clear, step-by-step workflow it could follow reliably.

Getting the agent to complete a 10-step, multi-tab process (find sources, find different sources, synthesize both) without "forgetting" its original goal was a huge challenge. It took dozens of refinements to craft the "Master Prompt" that was robust enough to handle different topics.

Accomplishments that I'm proud of

I'm incredibly proud of the "pivot" from V1 to V2. It's easy to make a summarizer; it's hard to automate insight. That leap—realizing the "Criticism Search" was the key—is the accomplishment.

I'm also proud of successfully "programming" an AI with natural language. The final V3 Master Prompt is a piece of code, and it proves that complex, multi-step automation is possible by "teaching" the AI a new skill, not just giving it a command.

What I learned

I learned that the future of automation isn't just about making us faster; it's about making us smarter. By automating the low-level task of data collection, this project frees up students to focus on the high-level task of creative, original thought. I learned how to "program" an AI agent, turning a high-level concept ("find me a thesis") into a concrete, executable set of instructions.

What's next for The Critical Insight & Gap-Finder

This workflow has huge potential. The next steps would be:

1. Saved Workflows: Work to save the "Master Prompt" as a one-click shortcut within Comet Browser.
2. Specialized Agents: Create new versions of the prompt that are specialized for different fields (e.g., "Legal Gap-Finder" that searches legal databases, or "Market Gap-Finder" for business students).
3. Expanded Output: Add a final step where the agent takes the generated "Insight Report" and automatically creates a 5-slide presentation outline based on it.

Built With

• comet-browser
• natural-language-prompt-engineering
• perplexity-ai