Recall
An AI-powered local search app that lets people find anything in their personal digital life — photos, scans, downloads, documents — using plain-language descriptions instead of filenames.
Everyday users and family archivists whose years of photos, scans, and downloads have become effectively unsearchable
- Plain-language search across local files ('Aunt Lina at the lake in 2019')
- Visual similarity, OCR, and date/scene heuristics across photos and PDFs
- Engineered to handle very large personal libraries (hundreds of thousands of files) without stalling
- Fully local and privacy-first, with optional encrypted sync as a paid add-on
The ripgrep-on-very-large-trees crash is the developer mirror of what every macOS, Windows, and Google Photos user already feels: built-in search stalls once a personal library grows past tens of thousands of items, leaving years of memories effectively unfindable.
Microsoft is committing to AI-powered personal file/photo search with Recall on Copilot+ PCs and Apple is similarly enhancing on-device AI, confirming real user pain around unsearchable personal libraries — but built-in options are already absorbing most of the demand.Microsoft's April 2025 Windows 11 Update: Recall, AI Search & the Future of Personal Computing ↗
Microsoft Recall (Windows), Apple Intelligence (macOS/iOS), and Google Photos search all directly target the same plain-language personal-search problem, and Big Tech has distribution + free bundling advantages — leaving very narrow whitespace for a third-party indie app.Microsoft's Controversial Recall AI Returns to Windows PCs ↗Privacy Concerns: Microsoft Recall and Apple Intelligence Auto Enablement ↗
Recall ships free on Copilot+ PCs and Apple Intelligence is bundled free with OS updates; charging $5–10/mo for a paid indie alternative against free, pre-installed, OS-integrated rivals is a very tough sell.Accelerating workplace productivity at Microsoft with Windows Recall ↗
Personal photo/document libraries only grow over a lifetime, and the underlying need (find old memories, scans, downloads) is permanent — not tied to any single trend cycle.
Doable but non-trivial: needs OCR, image/text embeddings, a vector index, on-device inference for privacy, plus cross-platform desktop indexing — meaningful engineering and ongoing model/storage costs.