Learning Brain

Inspiration

I save a lot of links to myself — research papers, GitHub repos, LinkedIn posts, articles, events, and tools — usually in WhatsApp. The problem is that later I remember the idea, but not the exact source.

Learning Brain turns that messy archive into a personal memory system I can query naturally.

What it does

Learning Brain:

• parses my WhatsApp self-chat and extracts saved links plus nearby notes
• uses Bright Data to fetch the underlying web content
• stores that material in Cognee Cloud as persistent structured memory
• uses an AWS Strands agent to retrieve and reason over what I saved
• lets me ask vague questions like “What was that LinkedIn post I saved about computer vision?”
• returns the original source URL
• can also remember a new URL live and make it immediately retrievable

How I built it

The pipeline is:

WhatsApp archive → Bright Data → Cognee Cloud → AWS Strands → FastAPI UI

I parsed 1,592 URLs from my WhatsApp self-chat. Bright Data retrieves the real page content, Cognee stores the memory and relationships, and Strands decides how to retrieve and use that memory.

I also built a lightweight FastAPI interface with two core actions:

• Ask my personal learning memory
• Remember a URL and add new knowledge live

What I learned

The most interesting part was the distinction between retrieval and memory.

Bright Data gives the agent reliable access to the underlying source material. Cognee turns that material into persistent memory. Strands provides the agent layer that decides what to recall and how to answer.

That separation made it possible to build something that feels less like search and more like remembering.

Challenges

The biggest challenges were:

• reliably fetching heterogeneous sources such as LinkedIn, GitHub, articles, and papers
• handling slow or blocked pages and timeouts
• preserving original URLs so retrieved memories remain actionable
• integrating Cognee Cloud memory cleanly with Strands
• making the full system reliable enough for a live demo in a very short hackathon window

The result is a personal learning brain that remembers the things I meant to come back to — even when I only remember a fragment of them.

Built With

• aws-bedrock
• aws-strands
• bright-data
• cognee
• css
• fastapi
• github
• html
• javascript
• python
• uvicorn