This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend.
What I Built
I built NextStep, a simple productivity app designed for a friend who often knows he has a lot to do, but gets stuck deciding what to do first.
The problem wasn't a lack of motivation. There were just too many things competing for attention.
So I built NextStep around one simple idea:
Don't figure out everything. Just figure out what to do next.
NextStep lets my friend quickly capture his thoughts and tasks, prioritize them by importance and timing, work through focused sprints, and track his progress throughout the day.
The most important feature I added for this challenge is AI Coach.
When my friend feels confused or overwhelmed about a goal, he can describe the situation in his own words. The AI Coach, powered by the open-weight Gemma model, turns that situation into three simple, actionable next steps.
Each AI-generated step can then be added directly to the task list.
This means the AI isn't just giving advice — it helps turn confusion into something my friend can actually do.
Demo
Live Demo:
https://newstep2.onrender.com
Try the AI Coach section and enter something like:
"I have an exam tomorrow and five chapters left. I don't know where to start."
The AI Coach will generate three practical next steps, which can be added directly to the NextStep task list.
Code
GitHub Repository:
https://github.com/rudra5002-code/newStep
The project is open source and contains the frontend, Express backend, AI Coach integration, and setup instructions.
How I Built It
I built NextStep as a lightweight web application using:
- HTML
- CSS
- JavaScript
- Node.js
- Express
- Hugging Face Inference Providers
- Google's Gemma 2 2B IT, an open-weight model
The core AI feature is the AI Coach.
The flow is:
- The user describes what's confusing or overwhelming.
- NextStep sends the situation to the Express backend.
- The backend sends the request to Gemma through the Hugging Face Inference API.
- Gemma is instructed to return exactly three short, practical next steps.
- NextStep displays those steps to the user.
- The user can add any generated step directly to their task list.
I deliberately made the AI output task-oriented rather than creating another general-purpose chatbot.
The goal is not to have a long conversation with AI.
The goal is to answer one question:
"What should I do next?"
I deployed the application on Render, with the AI request handled securely through the backend so the Hugging Face token is not exposed in the browser.
Why Does Open Innovation Matter?
Open innovation matters here because the AI model is part of the product experience, not just an external black box.
I used Gemma, an open-weight model, as the reasoning layer behind NextStep's AI Coach. That gives the project more flexibility than building the feature around a completely closed model.
Using an open-weight model makes it possible to think about future improvements such as:
- running the model locally
- experimenting with different model versions
- fine-tuning the model for productivity workflows
- changing the inference provider
- having more control over how the AI behaves
For a productivity tool dealing with someone's personal thoughts, goals, and plans, having more control over the AI stack is especially valuable.
Open models also make experimentation more accessible to individual developers. I was able to take an open-weight model like Gemma and turn it into a focused feature for a real person and a real problem.
Prize Categories
I am entering the following partner categories:
Best Use of Render
NextStep is deployed and running on Render as a live web application. The application uses a Node.js/Express backend to securely handle the AI Coach requests.
Live Demo:
https://newstep2.onrender.com
Best Use of Gemma
The AI Coach is powered by Google's Gemma 2 2B IT, an open-weight model.
Gemma is used as the core AI component that transforms an overwhelming situation into three practical next steps that the user can turn into tasks.
Top comments (0)