</sentinel>
</the problem>
Too many unreliable news sources. People can't track topics across outlets, languages, or verify what they read. Enterprise intelligence tools cost $500+/month, leaving students, journalists, and everyday people with no way to get verified, structured intelligence on any topic.
</my approach>
Built a full intelligence pipeline: Firecrawl scrapes global sources (including non-English with auto-translation), feeds them into Gemini 2.5 Flash for claim extraction and verification, and assembles structured reports with confidence scores. Iterated through 4+ rounds of low-fidelity testing to balance information density with readability.
</key features>
Topic tracking with push alerts
Subscribe to any topic and get notified when breaking developments happen. No manual re-searching.
AI reports with blind spots analysis
Every report includes what the coverage is missing, not just what it says. Confidence scores on individual claims.
Financial charts and timelines
Live market data from Finnhub, event timelines, and candlestick charts embedded directly in reports.
Cross-user SHA-256 cache
Duplicate queries hit a shared cache, cutting API costs 50-80% at scale.
Social feed and messaging
Unified feed with For You and Following modes, report likes/comments/follows, and full direct messaging between users.
Ask Sentinel AI chat
Conversational RAG over any report. Ask questions and get AI answers grounded in verified source data.
Audio briefings and daily stories
On-device TTS for hands-free listening. Gemini-generated daily story briefings summarize your tracked topics each morning.
</what i learned>
Balancing information density with readability is the hardest UX problem in intelligence tools. Too much data overwhelms; too little loses trust.
Self-chaining edge functions are a viable pattern for long-running background jobs on serverless platforms.
Shipping a v1 and iterating with real users beats perfecting behind closed doors. Most of the best features came from post-launch feedback loops.
On-device TTS (AVSpeechSynthesizer) eliminates API costs entirely for audio features and works offline — a better trade-off than cloud TTS at early stage.