jundot/omlx
omlx is an LLM inference server featuring continuous batching and SSD caching, optimized for Apple Silicon. It can be managed directly from the macOS menu bar.
Image: GitHub
The weekly · Week 20 · 11 May – 17 May 2026
AI coding agents: what changed
This week highlights the critical importance of AI agent security in production—spanning Base64 attack vectors and MCP implementations—alongside new developer tools like hermes-agent and specialized Figma design workflows.
6 picks
NousResearch/hermes-agent
Hermes Agent is designed to adapt and evolve alongside its user's needs. It is built to support continuous growth and learning.
Image: GitHub
Is Your Claude Code Safe From Base64?
This article explores potential security vulnerabilities in Claude Code when dealing with Base64 encoding. It discusses how AI agents might exploit such weaknesses in 2026.
Securing AI Agents in Production
This piece discusses strategies for securing AI agents deployed in production environments. It evaluates the strengths and weaknesses of MCP regarding AI agent security.
Building a Figma Design Agent
This describes the process of creating a Figma design agent for an Agentic AI Hackathon. The project involved leveraging AI for design automation.
Workshop: Building a Google Drive LINE Bot with Gemini CLI
This workshop provides hands-on experience in building an AI application using Gemini CLI and Official MCP. Participants will learn to launch a Google Drive LINE Bot from scratch.
Picked from public sources (GitHub, Hacker News, DEV, Lobsters, Reddit, Product Hunt) by our daily scout. Each summary uses only what the source itself says; GitHub stars are as checked on the date shown.