What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we follow →Needle2 is a 14MB agentic language model designed for phones, wearables, smart home devices, and robots.
Prime Agent is a self-improving reinforcement learning model agent.
Zero-Mem enables LLM agents to perform memory operations without consuming additional tokens.
EdotEnv is a Y Combinator startup providing reinforcement learning environments for quantitative trading to train large language models.
Prized is a tool for non-engineers to build secure internal tools.
Tokenless automatically switches between AI models to reduce costs for users.
Tines 3B provides safe workflow automation for software development environments.
Formally verified 3D CSG implementation with 93-line specification replaces 1000-line AI-generated code.
Screenpipe is a tool that records screen activity continuously to power AI agents with visual context.
Yorishiro is a macOS terminal application designed to host and run AI agents.
Large language models differ fundamentally from compilers and power tools in their design and operation.
Libretto PR agents automatically fix failing Playwright test scripts.
LM Studio Bionic is an AI agent for running open-source models.
Agnost AI extracts user feedback from agent conversations.
Oodle.ai offers agent trace storage at ten dollars per million traces.
Self-hosted voice AI agent for Asterisk/FreePBX telephone systems.
Juggler is an open-source GUI coding agent created by the JUCE framework creator.
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
