Reelbox - Turning the "Bookmark Black Hole" Into an AI Second Brain
Timeline
October 2026 (1-week sprint)
My role
Product Designer & Builder
(UX Architecture, GenAI Interaction Design, and Live Cloud Deployment)
Project type
Solo Build & Private Beta
(Built for personal daily use & tested with college friends)
Video bookmarking was broken. Design resources, Figma plugins, and tutorials got buried in Instagram and YouTube bookmark folders with zero search. Finding a saved reel meant endlessly scrolling through hundreds of unlabelled thumbnails. Most were never opened again. I designed Reelbox to make saved reels easily searchable whenever I actually need them.
I designed and built the app solo in a week with Antigravity, connecting a 1-tap Telegram save bot to an easily searchable web dashboard.
96% faster
Retrieval time (4m 12s → 9.2s)
Under 5 sec
To find any saved reel
the problem
The cost of zero search: saved reels get lost.
Zero search inside collections: Instagram and shorts only search public feeds. Saved folders have no keyword search, tag filters, or date sorting, leaving users stranded in an unsearchable grid.
Context-free video thumbnails: Thumbnails never show the tool name, link, or trick inside the reel. Remembering why you saved something requires scrubbing through 45-second clickbait videos.
The "Collector's Fallacy": Saving triggers an instant dopamine hit of productivity, but without an easy retrieval system, saved reels quickly pile up into an abandoned digital graveyard.
KEY DESIGN DECISIONS
Decision 1: 1-tap save over a native app
The Problem: Nobody wants to install another bookmarking app, copy-paste URLs, or manually pick folders mid-scroll. Every extra step kills mobile saving.
The Decision: I turned Telegram into the save engine. Users just tap native Share → Telegram → Reelbox. The bot automatically grabs the video, tags it, and syncs it to their private vault in under 2 seconds.
The Trade-Off ("Show the Fight"): The obvious choice was building an iOS/Android app. I killed the native app plan because piggybacking on an existing messenger eliminated install friction, login barriers, and app-switching fatigue completely.
01 · Native Share Sheet
No link copying. 1 tap to Telegram.
02 · Automated Ingestion
AI extracts category and summary.
03 · Synced to Vault
Instantly accessible on web.
Decision 2: Global search over manual browsing
The Problem: Instagram has zero search inside saved posts. Finding a specific tool or tutorial requires scrolling endlessly through hundreds of unlabelled thumbnails, guessing what's inside.
The Decision: I designed a keyboard-first global search (Ctrl + K) that indexes video titles, AI tags, and tools. Typing any keyword (e.g., "Framer", "Voice AI", "Portfolio") filters the entire vault in real time without refreshing.
The Trade-Off: Traditional tools (Notion, Raindrop) force users to navigate deep, rigid folder hierarchies. I prioritized fast global search because people remember what the tool was, not which folder they filed it into.
Decision 3: Text-first card over video playback
The Problem: When users revisit saved reels, they usually just want a specific tool name, link, or Figma tip. Having to re-watch and scrub through a 45-second video just to recall one detail causes high fatigue and drop-off.
The Decision: I designed scannable cards that surface the extracted tool name, clear headline, and category tags directly on the card surface. Users get the key takeaway in under 3 seconds without needing to press play.
The Trade-Off: The default pattern for video apps is an auto-playing video feed (like TikTok Web). I rejected auto-play because a personal reference vault should prioritize fast, calm reading over passive entertainment.
OUTCOME
Saved reels you can actually find.
Outcome 1:
96%
Retrieval time saved
(from 4m+ to under 10s)
Outcome 2:
4x
Higher revisit rate
(vs. 84% native abandon)
Outcome 3:
0%
Onboarding drop-off
(zero app installs)
Leraning
What building Reelbox taught me.
Product thinking before pixels:
Killing the native app plan for a 1-tap Telegram bot proved that reducing human friction matters more than drawing extra Figma screens.
AI-native velocity:
Using Antigravity to vibe-code and deploy live cloud software allowed me
to ship a complete working product in 7 days without engineering bottlenecks.
Designing for real production data:
I design systems that hold up under messy real-world data (scraped captions, missing thumbnails, shortcuts) so they thrive in production, not just in Figma.
Final design
Want to try Reelbox?
heyshobhitkeshari@gmail.com
+91 8840142636
MAIN