NIRA

Context

Standard chatbot explanations made it easy to feel familiar with system design, product management, and MBA concepts without exposing gaps or requiring transfer to a new problem.

Problem

I wanted an agent that would teach in the way I learn best: one focused question at a time, hints before answers, respectful challenges to unsupported reasoning, and proof of understanding before progress is recorded.

What I built

I built NIRA within the OpenAI environment using native scheduled tasks and context features, plus an Apple Notes MCP integration that saves session summaries, progress, weak areas, and future review topics.

How it works

1. 01Select the next concept each morning from the curriculum and previous progress.
2. 02Introduce it through a practical question or scenario rather than a complete explanation.
3. 03Evaluate my response, identify gaps, and ask one increasingly focused follow-up question at a time.
4. 04Adjust difficulty, offer hints, and use examples, counterexamples, and transfer questions to test understanding.
5. 05Mark a concept as understood only after I can explain it independently or apply it in a different scenario.
6. 06Save the session summary and next review plan to Apple Notes through MCP for future context.

Outcome

• Completed approximately 80% of the system-design curriculum.
• Completed approximately 50% of the product-management curriculum.
• Started approximately 5% of the introductory MBA curriculum.
• Created a consistent daily active-recall and progress-tracking routine.

What I learned

An educational agent should not confuse conversation with learning. Persistent state, adaptive questioning, application tests, and a clear completion rule make the workflow meaningfully different from a chatbot.