GURU AI for Android: Self-Evolving On-Device Assistant

What changed

The GURU AI project has released version v3.4.0, with the latest commit pushed on September 28, 2026. This release marks a recent update for the self-evolving AI assistant designed for Android devices. GURU is presented as a personal AI companion that aims to provide users with total control over their phone and network.

Key features highlighted include:

• On-device memory: Personal data is stored locally and is unique to the user.
• Automation tools: Capabilities to automate daily tasks and potentially earn money.
• Self-building tools: The AI is designed to build and improve its own tools.
• Privacy focus: All operations and data are confined to the user's device, with no data leaving the device.

The project is open-source, available as a sideloaded APK, and written primarily in Kotlin, with mentions of Rust in its topics. It supports the concept of 'bring-your-own-model' and aims for offline AI capabilities.

Why it matters for builders

For developers, GURU represents an exploration into building a truly on-device AI agent. The project's emphasis on self-evolution and local data processing offers a different paradigm compared to cloud-dependent AI solutions. Builders interested in offline AI, personal data privacy, and agent-based automation can find insights into architectural approaches and implementation strategies.

Practical impact

Android users can download the free, open-source GURU AI assistant as an APK. The project's recent release suggests ongoing development and support. Users seeking an AI companion that prioritizes privacy and offers a suite of tools for device control and automation may find GURU to be a viable option. The project's open-source nature also allows for community contribution and scrutiny.

Caveats and source limits

The repository metadata provides a high-level overview of GURU's features and recent release. Specific details regarding the AI models supported, the exact mechanisms of self-evolution, performance benchmarks, or the complexity of the tools it builds are not detailed in the provided source. The project has a low star count (3 stars) and no forks, indicating limited community adoption or visibility at this time. Further investigation into the project's codebase and documentation would be necessary to understand its technical depth and practical capabilities.