Save Instagram/TikTok via DM; retrieve by asking in natural language

Many active Instagram and TikTok users struggle to reliably save and later find short-form content they care about: saves, likes, and DMs become an unsearchable, ephemeral mess for the roughly 300 million "power social" users who frequently discover new content. This problem affects not only consumers trying to resurface a recipe, place, or fashion idea, but also creators and small brands who need organized archives of inbound content for reuse and analytics. The product would let users save posts by forwarding or DMing them into a private personal library that’s automatically indexed with multimodal AI—extracting captions, ingredients, places, faces, and scene metadata—and then retrieved via natural-language DMs or chat queries. The market opportunity is concrete now: short-form discovery is the primary feed on Instagram and TikTok, multimodal vision+language models make robust indexing practical, and the TAM is estimated at $18.0B (300M users × $5/mo ARPU × 12 months), with a Market Score of 95/100 and a Revenue Potential of 78/100. This can stand out by making privacy and ownership central—local-first or end-to-end encrypted storage, simple in-platform DM workflows, and high-quality, domain-tuned models for recipes, locations, and people—while offering creator-facing features like shareable collections and analytics. Key challenges are real: platform API access and policy risk, copyright and moderation liabilities, model accuracy on diverse UGC, and customer acquisition against a medium-competition landscape; the idea is worth pursuing if you have a clear technical plan for reliable multimodal indexing, legal strategy for platform integrations, and channels to reach power users who will pay the targeted ~$5/month.