Study Break

The Problem

Too many apps, none of them complete

Picture the setup: three browsers maxed with tabs. A flashcard app for memorization. A separate app for scheduling. Another for notes. Another for the calendar your professor half-filled out. A to-do list somewhere. Maybe a timer app for focus sessions. All of them requiring separate accounts, separate logins, separate paywalls for features that should have been free.

Now add ADHD to that equation. Every app switch is a context switch. Every context switch is an invitation to lose the thread. The tools designed to help you focus were actively fragmenting your attention.

And the content problem was worse. Most study apps offer generic material — different terminology, different emphasis, sometimes different conclusions entirely from what your professor taught. When the exam is based on Dr. Chen's slides, not a crowdsourced flashcard deck, generic content isn't just unhelpful. It's actively wasting your time.

Meanwhile, the AI revolution was happening and study tools were barely keeping up. Algorithms being updated years after the research was published. Paywalls on basic features. No awareness of your actual coursework. I was already running on a schedule that left zero room for inefficiency — and every one of these tools was adding friction instead of removing it.

Context

The schedule that demanded this

Wrestling. Six months a year, five days a week — plus private club training three additional days and competition most weekends for at least three of those months. You come home exhausted. You still have assignments due. You still have a dog who needs you. You still have a job.

My coaches were understanding — academics and feeding ourselves came first. But something instilled in me since martial arts as a teenager made being late to practice feel unacceptable. You wait respectfully until invited to join. You don't show up after that moment has passed. So the time I had for studying wasn't just limited — it was non-negotiable. Every minute had to count.

That's the real origin of Study Break. Not a hackathon idea. Not a business plan. A schedule that couldn't afford a single wasted study session and an ecosystem of tools that wasted them constantly.

The Catalyst

The professor calendar problem

This is the thing that pushed me over the edge. Professors manage their course calendars with wildly inconsistent levels of care. Some add every assignment to the LMS calendar. Some add half. Some add none and don't even publish a course outline. Some publish a syllabus with one set of due dates, post a different date in announcements, and have a third date on the learning management platform.

For students with jobs, athletics, or other commitments — people who need to plan their semester in advance — this is genuinely destabilizing. Adding some assignments but not others is worse than adding none, because it's misleading. You think you have a clear week. You don't. The missing entry costs you.

It cost me a few near-misses before I decided to solve it. I used to map out every due date manually, cross-referencing syllabi, announcements, and the LMS. Now the AI pulls them into the calendar automatically and I double-check its work. It also flags when a professor has given conflicting dates across sources — something I used to discover the hard way.

By the Numbers

The Solution

Everything in one place, finally

Study Break consolidates flashcards, scheduling, note processing, AI tutoring, campus navigation, analytics, and grade tracking into a single app. One account. One interface. One place where all your academic data lives together — so the AI can actually see connections between your classes that you might miss.

The Philosophy

AI that teaches, not replaces

I'd been thinking a lot about AI's role in education — what it helps, what it risks, where the line is. That shaped everything about how Study Break uses AI.

Most AI study tools optimize for output: generate the answer, pass the exam. Study Break optimizes for learning. The AI tutor has five modes, and none of them just hand you answers. Socratic mode asks strategic questions to guide you to discovery. Teach-Back mode makes you explain the concept, then evaluates your understanding. Adaptive mode reads your mastery levels and adjusts — direct explanation when you're lost, probing questions when you're close, nuance and edge cases when you've got it.

There's a difference between a tool that helps you learn and a tool that learns for you. Study Break stays on the right side of that line.

Design Constraint

Built by an ADHD brain, for ADHD brains

Every design decision in Study Break passed through a filter most apps never consider: will this survive a brain that's actively looking for a reason to context-switch?

Fewer tabs, fewer accounts, fewer logins. That's not a convenience feature — it's a survival feature. Every external app is an exit ramp. Every new browser tab is a trapdoor. Consolidating everything into one place isn't about aesthetics. It's about keeping the door closed long enough to actually study.

Ambient sounds — white noise, rain, café hum — because silence is its own kind of distraction. Pomodoro timers with structured breaks because "just keep going" is not a strategy that works for every brain. Speed-running detection that flags when you're flipping cards too fast — catching the moment your eyes are moving but your mind checked out. Interleaved practice that deliberately mixes subjects, because monotony is where ADHD focus goes to die.

These aren't accessibility features bolted on after the fact. They're the architecture. The app was designed this way because it had to be — for the person building it to use it.

The Hard Parts

What made this difficult

• FSRS-5 spaced repetition from current research The algorithm that decides when you see a card again matters more than anything else in a flashcard app. Study Break implements FSRS-5 — the latest generation, with 19 optimized parameters and a full state machine tracking stability, difficulty, and retrievability per card. Five review formats (standard, reversed, cloze, free recall, multiple choice) because recognition and recall are different cognitive processes.
• Parsing the unparseable Syllabi have no standard format. Due dates are embedded in prose, buried in tables, contradicted by announcements. Late policies range from "10% per day" to "tiered: 1–2 days 5%, 3+ days zero credit" to "don't bother." The AI extracts structured data from all of it — penalty types, grace periods, maximum late days — and normalizes it into something the catch-up planner can actually calculate against.
• AI context that actually knows you The tutor doesn't just answer questions. It carries per-class memory — your struggles, your misconceptions, your upcoming deadlines, your mastery levels ranked by Bloom's taxonomy. Every conversation builds on the last. That required careful context management: injecting student state, class documents, and conversation history without blowing past token limits.
• Full offline support with sync Students study on buses, in gyms, in buildings with terrible WiFi. Study Break runs entirely offline via IndexedDB with background sync when connectivity returns. Not read-cache. Full offline — create cards, complete reviews, log sessions, all without a connection.
• Keeping it free Students are broke. Putting paywalls on basic study features felt wrong. The free tier includes everything except AI messages. OpenStreetMap instead of Google Maps. Supabase instead of Firebase. Every architectural decision was filtered through "can I afford to give this away?" We'll see if I missed something that starts costing money, but the intent is to keep as much free as possible.

Architecture

How it's built

React frontend with TypeScript, built on Vite. 30+ Radix UI primitives for accessibility out of the box (WCAG 2.1 AA). Express backend with Drizzle ORM over Supabase PostgreSQL. Claude AI powers tutoring, flashcard generation, document summarization, late policy parsing, and catch-up planning. Capacitor wraps the whole thing for iOS and Android.

The AI integration runs through 37 service modules — each handling a discrete capability. Document processing is async and non-blocking: text extraction happens immediately, summarization and formula detection run in the background. The tutor's context injection layers student state, class data, mastery levels, and conversation memory into every prompt.

196 automated tests. Sentry for crash reporting. Axe-core accessibility testing in CI. The campus map runs on Leaflet and OpenStreetMap because Google Maps API pricing is not student-friendly — and honestly, most campus maps could use the upgrade anyway.

What I Learned

The honest retrospective

The biggest lesson was that consolidation is the feature. Not any single tool — not the flashcards, not the AI, not the calendar. The value is that everything lives together. When all your academic data is in one place, the AI can find interdisciplinary connections your professors never mentioned. The calendar can warn you about deadline collisions before they happen. The catch-up planner can read your late policies and your grade weights simultaneously and tell you where your time is best spent.

None of that works when your data is scattered across six apps. Fragmentation isn't just inconvenient. It's a ceiling on how smart your tools can be.

The other lesson was personal. I spent years thinking my brain was the problem — that I couldn't focus, couldn't plan, couldn't keep track of things the way other students seemed to. Turns out the tools were the problem. When the environment is designed for how your brain actually works, the brain works fine.

Study Break exists because advancements in AI finally made it possible for one person to build what used to require a team. I took a bunch of services that usually require multiple sites, multiple accounts, and multiple paywalls, and made as much of it free as I could. For every student running on a schedule that can't afford a single wasted session — this is what I needed. I'm really excited about it.