Un Segundo Intento

An AI-powered study coach that turns focus patterns, energy levels, goals, and completed work blocks into personalized guidance for building consistent and sustainable study habits.

Comment

Inspiration

Un Segundo Intento started with a simple idea: students do not need another app that expects them to be perfectly disciplined every day. Sometimes you lose focus, motivation, or consistency. What matters is having a way to start again.

Most productivity tools only tell you how many hours you studied. They do not explain why you are struggling or what you could change. I wanted to combine study tracking, accountability, and personalized guidance in one place, helping students build routines they can actually maintain.

What it does

Un Segundo Intento is a platform for students and exam candidates that makes studying easier to organize, understand, and improve.

The existing platform includes:

  • Focused study blocks.
  • Energy and concentration tracking.
  • Streaks and progress indicators.
  • Personalized study profiles.
  • Rankings, groups, and shared challenges.
  • Accountability and validation systems.

For OpenAI Build Week, I am adding an AI study coach powered by GPT-5.6. It uses each student’s profile, recent study blocks, energy levels, goals, and consistency to give them personalized guidance and one realistic action they can take next.

I do not want it to feel like a generic chatbot. The coach should help students understand when they focus best, whether their goals are realistic, what is affecting their consistency, and what they can change in their next study session.

How we built it

The existing platform was built with Next.js and Supabase. Next.js provides the application interface and PWA experience, while Supabase manages authentication, the PostgreSQL database, and user activity data.

For the Build Week extension, I am creating a controlled context pipeline:

  1. Select only the study data needed for the recommendation.
  2. Turn it into structured context without including unnecessary personal information.
  3. Send that context to GPT-5.6 with clear instructions and limits.
  4. Validate the structured response.
  5. Show the student a short explanation and a practical next step.

I am using Codex throughout the process to understand my existing codebase, plan the architecture, implement the data and interface layers, test different situations, and document the new functionality.

Because Un Segundo Intento already existed before Build Week, I am separating all the new work through dated commits, clear documentation, and identifiable Codex sessions.

Challenges we ran into

The main challenge is making the advice feel truly personal without making assumptions that the available data cannot support. The coach needs enough context to be useful, but it should not use more personal information than necessary.

Another challenge is finding the right balance between accountability and sustainability. Rankings, streaks, and recommendations can motivate students, but they should not encourage exhaustion, impossible goals, or unhealthy comparison with other people.

I also need to make the AI coach feel like a real part of Un Segundo Intento instead of a separate chat window. Its recommendations should be connected to what the student is actually doing and give them something clear they can do inside the platform.

Accomplishments that we're proud of

I am proud to have built a real platform that connects study sessions, personal patterns, progress, and accountability in one place.

I am also proud that the Build Week extension uses AI to solve a specific problem. GPT-5.6 is not being added just to say that the project uses AI. It receives structured study data so its recommendations can be useful in the student’s real routine.

Finally, I have kept a clear separation between the original platform and everything created during Build Week. This makes the development process more transparent and easier to understand.

What we learned

Building Un Segundo Intento has taught me that recording study time is only the beginning. The difficult part is turning that information into feedback that arrives at the right moment, is easy to understand, and gives the student something useful to do.

I have also learned that building a useful AI feature requires much more than writing a prompt. It needs carefully selected context, structured responses, validation, privacy limits, and a clear idea of the decision the AI is helping the student make.

Codex has shown me how an AI development agent can help me move faster through an existing codebase while still leaving the important product and architecture decisions in my hands.

What's next for Un Segundo Intento

My immediate priority is to complete the AI study coach, test its recommendations with different study profiles, and improve the experience using feedback from real students.

Later, I want to add weekly progress reviews, detect inconsistent or unsustainable study patterns earlier, explain recommendations more clearly, and create better tools for study groups and mentors.

My long-term goal is for Un Segundo Intento to become more than a study timer. I want it to help each student understand how they study and build a routine they can realistically maintain.

Built With

  • codex
  • css3
  • ffmpeg
  • gpt-5.6
  • html5
  • javascript
  • node.js
  • openai-api
  • postgresql
  • progressive-web-app
  • puppeteer
  • resend
  • service-worker
  • stripe
  • stripe-billing
  • stripe-checkout
  • stripe-webhooks
  • stylelint
  • supabase
  • supabase-auth
  • supabase-storage
  • vercel
  • vercel-cron-jobs
  • vercel-functions
  • web-app-manifest
Share this project:

Updates