I built Shoof Aflam — an Arabic streaming guide with 13,400+ titles across 17 platforms — in under a month. The secret? I coded the entire thing with Claude (Anthropic's AI), pair-programming every component from data pipelines to SEO optimization. Here is the real, honest story.
How It Started
I had a simple problem: there was no Arabic equivalent of JustWatch. If you wanted to know where to legally watch an Arabic movie or series, you had to check every platform manually. No centralized guide existed.
Instead of spending months planning architecture, I opened Claude Code and started building. The entire project — from first commit to 13,400 indexed titles — took about 4 weeks of intensive pair-programming sessions.
The Stack (Built With Claude)
- Next.js 16 with full static export — ~7,000 pre-rendered HTML pages
- TypeScript 5 strict mode
- Tailwind CSS 4 for RTL-first Arabic design
- Python 3 scripts for TMDB/YouTube/Wikipedia data collection
- 129 Vitest tests enforced as pre-build gate
- Nginx + Cloudflare on a single Hetzner VPS (not Vercel)
Every line of code was written in collaboration with Claude. I described what I wanted, Claude wrote the implementation, I reviewed and iterated. Some components went through 5-6 revisions in a single session.
What Claude Actually Built
Here is what we built together, honestly:
Data pipeline — 10+ Python scripts that:
- Mass-discover Arabic titles from TMDB API (went from 424 to 13,400+ titles)
- Enrich with cast, crew, trailers, ratings, production data
- Fetch Arabic overviews and translate English ones via Google Translate
- Pull Wikipedia Arabic extracts for cultural context
- Generate unique FAQ entries from real title data (98.6% unique across 64,800 entries)
- Collect YouTube review comments with optimized API quota management
SEO system — Claude ran multiple audit rounds and fixed everything:
- Schema.org JSON-LD for 11 entity types with sameAs links to IMDB/TMDB
- Dynamic title tags, featured snippet blocks, OG meta
- Thin content detection and noindexing (went from 43% thin to 0%)
- Content enrichment bringing pages from ~50 words to ~190 words average
- HTTP/2, lazy-loaded search index, LCP optimization
Outreach automation — A complete system running on the VPS:
- Automated discovery of guest post blogs
- Email outreach with personalized pitches
- Backlink monitoring with Telegram notifications
- All running via cron, zero manual intervention
The Data: Real Numbers
| Metric | Value |
|---|---|
| Total titles | 13,426 |
| Indexed pages | 11,766 |
| With Arabic overview | 98% |
| Production data coverage | 99.8% |
| Wikipedia facts | 1,400+ |
| Unique FAQ entries | 64,813 |
| Vitest tests | 129 |
| Static pages generated | ~7,000 |
| Build time | ~13 seconds |
Lessons From AI Pair-Programming
What worked great:
- Claude excels at generating boilerplate, test suites, and data processing scripts
- Running multiple Claude agents in parallel (5-6 at once) for SEO audits was incredibly productive
- Claude caught issues I would have missed — like uploadDate missing from VideoObject schema, or FAQ answers being 88.7% duplicates
What needed human judgment:
- Deciding what to noindex vs keep (content quality is subjective)
- Evaluating whether outreach emails sound natural
- Choosing which data to trust (TMDB has gaps for Arabic content)
- Strategic decisions: which platforms to target, what content adds real value
What Claude got wrong:
- Initially generated overview descriptions that were too templated
- Some SEO recommendations conflicted with AdSense compliance
- Automated provider refresh found zero MENA data (TMDB/JustWatch dont cover Arabic streaming)
- A few agents tried to build spam-like systems that I had to reject
The Hardest Part: Platform Data
The biggest unsolved problem: 80% of titles have no streaming platform data. TMDB and JustWatch have almost zero coverage for Arabic content in MENA regions. The data simply doesnt exist in any API. Getting real platform coverage requires scraping individual platform catalogs — Shahid, WatchIT, OSN+ — which is a separate engineering challenge.
SEO Results (Honest Assessment)
After the full SEO overhaul:
- Google organic traffic: still low (~10/day) — the site is new
- Technical SEO: excellent (all schema valid, no errors in GSC)
- Content quality: dramatically improved but still 58% of pages under 200 words
- The real test will be over the next 2-3 months as Google recrawls
Try It
- Arabic Streaming Report — data visualizations
- Platform Comparison — Shahid vs Netflix vs OSN+
- Telegram Bot — search any Arabic title
The entire codebase was built with Claude in about 4 weeks. If you are considering AI pair-programming for a side project, my advice: it works remarkably well for data-heavy, SEO-focused static sites. Just keep your human judgment engaged for quality and strategic decisions.
Alex Nitrovic builds shoofaflam.tv, the largest Arabic streaming guide. The project was built entirely with Claude (Anthropic) as an AI pair-programming partner.
Top comments (0)