InsightAI

Inspiration

With the growing volume of information available online, many critical technical, ethical, and human perspectives often remain underreported or inadequately explored, even across multiple reputable sources. InsightAI was inspired by the need to systematically identify these missing angles and present information that is truly relevant to people, rather than repeating what is already widely covered.

What it does

InsightAI is an autonomous intelligence agent designed to create accurate and relevant content. It analyzes numerous global sources and synthesizes a report that highlights systematically underrepresented aspects of a topic. By identifying missing perspectives, the application provides users with insights that go beyond surface-level reporting and focus on overlooked technical, ethical, and human dimensions.

How we built it

The application leverages a multi-agent architecture powered by Google Gemini 3 models.
The system first uses the Tavily API to extract raw content from multiple global sources. Google Gemini 3 Flash is then used to create topic maps based on the content reported in the analyzed sources. Its low latency, cost efficiency, and higher throughput make it ideal for this task.

Following this, the application consolidates all the topic maps into a unified research context. This context is used to perform investigative analysis and identify all missing aspects across the sources. The same information is then used to create a comprehensive report highlighting information that is truly relevant to people. Google Gemini 3 Flash is selected for its strong long-context reasoning abilities, empowering cross-document comparison and identification of implicit gaps and secondary implications.

Challenges we ran into

One of the key challenges was enabling effective cross-document reasoning while maintaining low latency and cost efficiency. Identifying implicit gaps and underrepresented perspectives across diverse sources required careful consolidation of topic maps without introducing redundancy or bias.

Accomplishments that we're proud of

We built an autonomous system capable of synthesizing insights from multiple reputable sources and systematically identifying missing perspectives. The multi-agent architecture allowed us to generate a unified research context and produce reports that highlight information often overlooked by conventional analysis.

What we learned

We learned how multi-agent architectures combined with long-context reasoning models can significantly enhance investigative content analysis. We also gained insights into balancing reasoning depth, throughput, and cost efficiency at scale.

What's next for InsightAI

Next, we plan to enhance InsightAI by expanding source coverage, improving topic map granularity, and refining how missing perspectives are prioritized, making the system even more effective at surfacing insights that matter most.

Built With

• css
• gemini3
• html
• javascript
• python
• regex
• tailwind
• tavily
• tenacity
• vercel