Approach

Inspiration The idea for Approach came from a simple observation: sometimes students are not struggling because they are incapable of doing a task—they are struggling to approach the task in the first place. A mathematics teacher once described this beautifully: the studying itself belongs to the child, but helping them reach that work and reducing their fear, doubt, or hesitation belongs to the teacher. We wanted to turn that philosophy into a digital learning companion. Importantly, this problem is not limited to neurodivergent students. Procrastination, task avoidance, overwhelm, uncertainty, low motivation, and fear of failure are extremely common among students and young people. Approach is designed to support neurodivergent learners particularly well while remaining useful to anyone who gets stuck before starting. What it does Approach asks the learner what is making a task difficult right now—for example: “I don't know where to start.” “It feels overwhelming.” “I'm anxious about it.” “I don't have the energy.” “I've been avoiding it.” “The deadline is making it worse.” Or they can describe something completely different. The system then uses AI to understand the barrier and suggest one small, practical first step toward the actual task. The goal isn't to tell students to “just do it,” nor is it to encourage indefinite avoidance. Approach reduces the distance between the learner and the work. How we built it We built Approach as a lightweight web application using Next.js, React, TypeScript, Tailwind CSS, and Gemini AI. The frontend provides a calm, low-friction interaction where the learner can identify their current barrier without being judged. The AI layer receives that barrier through a server-side API route and generates contextual guidance. The API key remains server-side rather than being exposed to the browser. Our core interaction is deliberately simple: Student ↓ Identifies the barrier ↓ Approach understands the barrier ↓ AI suggests one manageable first step ↓ Student approaches the actual task We also designed the system around an important principle: support should adapt to the student's current state without turning temporary difficulty into an excuse to abandon important work. Challenges we ran into The biggest challenge was building a meaningful intervention without making Approach another productivity tool that simply tells students what to do. We had to carefully balance: empathy with accountability; reducing overwhelm without encouraging avoidance; supporting neurodivergent learners without treating them as fundamentally different from everyone else; giving useful AI guidance without pretending the AI can diagnose a student's condition; keeping the interaction extremely simple while still being meaningful. We also faced significant time constraints during development, including debugging the Next.js API routing and Gemini integration immediately before submission. Accomplishments that we're proud of We are proud that Approach focuses on a small but often overlooked point in the learning journey: the moment before the student starts. Rather than replacing the teacher or doing the student's work, Approach tries to play the role of a supportive bridge between the learner and the task. We are especially proud of the “Something else…” option. Students do not always fit predefined categories, and the system should listen when our assumptions are wrong. Most importantly, the product doesn't define success as “the AI answered the student.” Success means the student is a little closer to doing the work themselves. What we learned We learned that educational AI does not necessarily need to become more complicated to become more useful. A relatively small intervention can be powerful when it happens at the right moment. We also learned that accessibility should not mean designing only for a particular diagnostic category. Many supposedly “neurodivergent” barriers—difficulty starting, overwhelm, procrastination, uncertainty, fear of failure—are part of ordinary human learning too. The challenge is therefore not simply: “How can AI teach the student?” It is also: “How can AI help the student become ready to learn?” What's next for Approach The next version of Approach will move beyond identifying a barrier and suggesting a first step. We want Approach to become a closed learning loop: Approach → Start → Work → Reflect → Adapt → Approach again. Future versions could: follow up after the suggested first step; recognize whether the student actually progressed; adapt the size and type of the next step; incorporate deadlines and task urgency more intelligently; provide stronger teacher/parent collaboration while preserving student autonomy; support richer accessibility preferences; measure whether Approach actually improves task initiation, persistence, and learning outcomes. Ultimately, we don't want to build an AI that does students' work for them. We want to build an AI that helps students get close enough to the work that they can do it themselves.