GitHub 2026-04-10 Weekly Trending Projects

AI Edge Gallery is an on-device offline generative-AI sandbox for mobile, supporting local Gemma 4 inference, model management, and multimodal interaction—suitable for privacy-sensitive development and performance evaluation.

oh-my-codex layers standardized sessions, reusable skills, and a team runtime on top of Codex (without replacing it), suited for teams and individuals who want workflow-driven AI agent operations with durable, auditable state.

OpenScreen is an Electron-based free open-source screen recorder and demo maker offering recording, zooms, annotations and multi-resolution export—well suited for product demos and tutorials; note system-audio compatibility issues and incomplete repository metadata.

Hermes Agent is a server-deployable autonomous AI agent offering a terminal-first interface, multi-platform messaging gateway, persistent memory and skills hub, sandboxed terminals, and scheduling—suitable for long-running, auditable, and extensible automation in engineering and operations.

LiteRT-LM is Google's edge-focused large-model inference framework emphasizing multimodality and hardware acceleration, suited for deploying high-performance offline or near-edge inference on mobile, desktop, and embedded devices.

DeepTutor centers on an agent-native architecture that combines TutorBots, persistent memory, and RAG knowledge bases to deliver an extensible personalized tutoring and collaborative workspace—well suited for autonomous learning, teaching automation, and knowledge-driven conversational education products.

Onyx is a self-hosted AI chat and RAG platform for teams, compatible with diverse LLMs and offering extensive connectors and deployment options, aimed at enterprise use with ops capability.

TimesFM, released by Google Research, is a decoder-only time-series foundation model that strengthens long-context forecasting and fine-tuning capabilities; it suits research and enterprise univariate forecasting use cases but demands substantial compute/memory and dependency engineering for production deployment and compatibility.

PersonaPlex, built on Moshi and the Helium backbone, delivers low‑latency full‑duplex speech dialogue with combined text‑role and audio voice control, suited for research, real‑time interactions, and customer‑service prototyping and evaluation.

A hands‑on, example‑driven guide to Claude Code offering 10 modules and copy‑ready slash commands, hooks, skills and MCP configs to help teams build automated workflows within hours — however, repository metadata (commits/contributors/license) is inconsistent with the README, so verify origin and maintenance before production use.

Condenses Andrej Karpathy's observations about LLM coding pitfalls into four installable CLAUDE.md principles—explicit assumptions, simplicity-first, surgical changes, and goal-driven execution—helping teams limit unnecessary edits and make LLM-driven changes verifiable.