ScreenMind appeared on Hacker News as a local privacy-first Microsoft Recall alternative built with Gemma 4. The project is hosted at github.com/ayushh0110/ScreenMind and received 11 points with 2 comments in its first discussion thread.
Model: Gemma 4 | Deployment: Local only | Focus: Screen capture + semantic search | License: Open source
What It Is and How It Works
ScreenMind runs entirely on the user's machine. It captures screenshots at intervals, stores embeddings locally, and lets users search past screen content through natural language queries. Gemma 4 handles the embedding and retrieval steps without sending data to external servers.
The system avoids the cloud dependency that Microsoft Recall introduced. All processing and storage remain on-device.
Benchmarks and Technical Details
No public benchmark numbers were shared in the initial post. The HN thread contains only the project link and a short description. Early visitors noted the focus on local execution and privacy controls rather than speed metrics.
How to Try It
Clone the repository from the provided GitHub link and follow the README instructions for installation. The project uses standard local LLM setup steps with Gemma 4 weights downloaded separately.
Users need a machine with sufficient RAM to run Gemma 4 inference alongside the capture service. No API keys or accounts are required.
Pros and Cons
- Runs fully offline with no data leaving the device
- Uses open-source Gemma 4 for embeddings
Simple semantic search over personal screen history
Limited community feedback so far (only 2 comments on HN)
No published latency or accuracy numbers
Requires manual setup compared with commercial Recall builds
Alternatives and Comparisons
Microsoft Recall stores encrypted snapshots in the cloud with optional local processing. Other local tools such as Rewind.ai require paid subscriptions and macOS-specific hardware.
| Feature | ScreenMind | Microsoft Recall | Rewind.ai |
|---|---|---|---|
| Deployment | Local | Cloud + local | Local + paid |
| Model | Gemma 4 | Proprietary | Proprietary |
| Data sharing | None | Optional | None |
| HN visibility | 11 points | N/A | N/A |
Who Should Use This
Developers and researchers who need on-device screen search without external accounts will find ScreenMind relevant. Users requiring enterprise support, polished installers, or published performance benchmarks should wait for further releases.
Bottom line: ScreenMind provides the first open local alternative to Recall that explicitly keeps all data and inference on the user's hardware.
The project remains early-stage, with community adoption depending on additional contributors adding benchmarks and easier installation paths.
Top comments (0)