Arkitekto: Philippine Architect Board Exam Reviewer

February 1, 2026

The problem

Aspiring architects in the Philippines face a difficult board exam: the Architect Licensure Examination (ALE), administered by the Professional Regulation Commission (PRC). Pass rates have historically ranged from 40-65%, with the January 2026 sitting reaching 82%. Yet study resources are scattered. Reviewers rely on outdated photocopied materials, expensive review centers, or random question compilations shared in Facebook groups.

Nowhere could an examinee practice against structured questions, see which subjects they were weak in, and study without wasting time. The distance between the tools available and what students actually needed was enormous.

What I built

Arkitekto is an online review platform covering every ALE subject, with 8,000+ practice questions behind it.

Core features

Practice by subject, with feedback and an explanation the moment you answer
Timed mock exams that follow the real ALE format, scored and reviewable afterwards
Spaced repetition on the SM-2 algorithm, scheduling questions by retention the way Anki does
Focus sessions aimed at whichever subjects you keep getting wrong
A flashcard mode for fast passes through a topic
Progress tracking: streaks, subject mastery levels, performance analytics, achievement badges
Study groups, so people can work through questions together
Discussion threads attached to individual questions
Bookmarks for the questions worth coming back to
An admin dashboard for content moderation, users, questions, and analytics

Tech stack

Layer	Technology	Why
Framework	Next.js 16	App Router, Server Components, React 19
Database	Supabase	PostgreSQL with RLS, Auth, real-time
Styling	Tailwind CSS 4	Utility first, design tokens
Components	shadcn/ui	Accessible, customizable primitives
Validation	Zod	Runtime schema validation
Hosting	Vercel	Edge caching, fast deploys

Architecture

Data flow for spaced repetition

Question schema

What building this taught me

1. Content quality is everything in EdTech

I started with 7,000+ questions and thought the platform was ready. Then users started reporting inaccuracies. A wrong answer key on one question destroys trust in the entire platform.

I ended up running 12 rounds of automated and manual review across the entire question bank. AI-assisted review helped catch issues at scale, but every flagged question still needed human verification against authoritative sources like the National Building Code and RA 9266.

A smaller set of verified questions turned out to be worth more than a large set nobody had checked.

2. Spaced repetition needs careful tuning

The SM-2 algorithm is well documented, but implementing it for exam preparation is different from language learning. Architecture exam questions aren't binary recall. They require understanding concepts across multiple subjects.

I had to adjust intervals and ease factors to account for the fact that missing a question about structural design doesn't mean you forgot it. You might have misunderstood the question or confused similar building codes.

Resources:

3. Row Level Security changes how you think about data

Supabase RLS means every query is automatically scoped to the authenticated user. This is powerful, but it forces you to design your schema around access patterns from day one.

Every table needs a user_id column. Every policy needs to be tested. And when you need cross-user queries (like study group leaderboards), you need database functions that bypass RLS with SECURITY DEFINER.

-- RLS policy: users can only see their own stats
CREATE POLICY "Users can view own stats"
  ON sc_user_question_stats
  FOR SELECT
  USING (user_id = auth.uid());

4. Niche audiences are deeply engaged

Building for Filipino architecture students means a small but passionate user base. Every feature request comes with genuine context. Users tell you exactly what they need because the exam is real and the stakes are high.

This taught me that building for a niche is easier than building for everyone. The feedback loop is tighter, the community is supportive, and feature prioritization is obvious because the goal is clear: pass the board exam.

The bigger realization

EdTech turned out not to be about technology at all. The hard problems here were content quality, whether the thing actually teaches, and understanding how people study when an exam is coming.

A much simpler platform with plain quizzes would have served the same purpose. Spaced repetition, analytics, and study groups only matter because the questions underneath them are correct. Good technology makes good content go further, and it does nothing for bad content.

What I would do differently

Invest in content review infrastructure from the very beginning. I built the platform first and treated content quality as a later concern. That meant months of retroactive review scripts and manual verification.

If I started over, I'd build a review pipeline first: a system where every question goes through automated checks and human verification before it ever reaches a student.