Brainch

An easy path to track your mind

Comment

Inspiration

What if you could power your chat with yourself to boost your productivity and even free your mind? What if just a few messages could start building a huge, fully personalized tool that stores what your brain doesn’t need to carry around.

Studies suggest that writing down the things you need to remember can significantly improve your rest and, as a result, your health and wellbeing.

That’s what inspired Brainch: a powerful tool that builds a brain-like structure from your daily needs.

The name comes from combining brain and branch, because its structure represents thoughts the way your mind maps them: as interconnected branches.

What it does

Brainch gives you a place to write down your thoughts, represent them as your brain would map them, and gather relevant information to generate suggestions and analysis about what you could do to hit your goals, deadlines, and plans.

The main characteristic of Brainch is how effortlessly simple it is: using it doesn’t even require a decision beyond thinking “I want to note this”—and instantly having it inside your brain… more precisely, your digital brain.

How we built it

Our early prototype has been built with:

  • FastAPI for the backend
  • React + Tailwind for the frontend
  • For basic tag generation, we’re using a local custom model, running a dockerized Ollama instance based on Llama 3.2
  • We also accelerated prototype development using Lovable and Codex

Challenges we ran into

  • We 'lost' our frontend implementation because it was in an individual repo cloned inside the main repository which eventually leaded to a loss of data less than 6h before the deadline
  • Our practical AI content did not match our goal

Accomplishments that we're proud of

  • Our goal was to take any challenge we decided to take and give it an original, different twist—and we did.
  • We worked through frustration to ship an MVP, even when getting anything on time felt nearly impossible.
  • The models we found for free that we were able to execute locally were not good enough to follow formats so they are not generating tags as expected

What we learned

  • We learned what embeddings are, and discussed multiple mathematical approaches to solve the problem in an optimal way.
  • We learned that hackathons are more about having something that shows your idea than making everything perfect.

What's next for Brainch

  • We’re definitely going to keep thinking about how to algorithmically sort new data optimally, grouping different things the way the brain would.
  • Our main objective is to build a model that can relate different messages without any specified context, using only the order the user sent them, the timestamp, and previously sorted elements.

Licensing

We've chosen GNU GPL-3.0 as our license because we don’t want this project to be just a reference to other repos: we want the real code inside this repo, so when someone clones it they get the full MVP. GPL-3.0 fits that idea because if someone uses, changes, and shares our code, they must keep it open too instead of turning it into a closed project.

Built With

Share this project:

Updates