CogniStream AI

• 

  Custom Socratic System Instructions for high reasoning output.

CogniStream AI: Logic-First Learning

Inspiration

Students often get overwhelmed by massive academic videos and dense PDFs. Traditional AI tools typically provide flat summaries that encourage rote memorization. My inspiration was to build CogniStream AI, which uses Gemini 3 Flash to transform these resources into interactive Logic Maps.

How I built it

• Engine: I utilized Google AI Studio to access the Gemini 3 Flash model, leveraging its 1M+ token context window.
• System Design: I engineered custom System Instructions to enforce Socratic reasoning, ensuring the AI deconstructs the 'Why' behind concepts.
• Optimization: I configured the model parameters to balance academic precision with highreasoning pedagogical outputs.

Challenges I faced

The primary challenge was ensuring the model didn't just summarize. By refining the System Instructions, I forced the model to generate hierarchical logic bridges. I also managed to resolve export issues by optimizing the video demo workflow.

What I learned

I learned the power of Prompt Engineering and how the Gemini 3 Flash context window can process 10,000+ tokens without losing coherence.

Built With

• ai
• api
• gemini
• google
• language
• large
• models
• studio