Tag
Built an agent loop where an LLM writes Python code to generate articulated 3D CAD models, with the ability to pin a part for editing.
A no-framework walkthrough for building a simple AI agent by hand, focusing on the core loop of goal, tool call, observe, and decide, with practical tips on prompt design, tool exposure, step limits, and logging.
A developer built an AI agent in 100 lines of Common Lisp, with the only tool being eval. The model executes code through a recursive agent loop and restores skills via a persisted transcript, showcasing Lisp's unique advantage as an AI language.
The author reflects on learning Lisp for symbolic AI 25 years ago and demonstrates how a modern AI agent loop can be implemented in just 100 lines of Common Lisp, emphasizing the elegance of recursion.
Jerry Liu announces LiteParse, a fast and accurate file parser for agent loops, now integrated with Vercel's Eve framework.
Aletheia is an open-source agent loop designed for investigating questions where answers cannot be easily verified, such as vendor credibility or financial health. It maintains an explicit belief state, chooses searches to maximize information gain, and can stop without forcing a conclusion when evidence is insufficient.
Jerry Liu presents a framework for document parsing across accuracy, cost, and latency tradeoffs, introducing LiteParse as an open-source, low-latency parsing tool for AI agent loops, along with LlamaParse for high-accuracy modes.
Google released an 8-minute tutorial that walks you through building a minimal Agent loop, emphasizing understanding concepts through practice rather than just managing prompts.
This post explains how to create an automated feedback loop for AI agents to iteratively improve their skills, using computer use and an observer skill to evaluate and update the skill code.
A guide on building a reusable Claude Code Agent loop that can be pointed at different tasks like bug fixing, speed optimization, or cost reduction by swapping check scripts.
agent-browser is a CLI tool for browser automation designed for AI agents, using compact text output and ref-based element selection to minimize token usage. The post also highlights three other tools—portless, emulate, and ai-cli—for improving agent loop efficiency.
The article argues that the agent loop is fundamentally the ReAct pattern, and that current tool-use APIs already implement this mechanism.
The article introduces how Matt Van Horn achieved 360 PRs in 30 days by designing an agent loop rather than writing prompts, and details his three tools and eight LAW rules, emphasizing the shift from prompt writer to loop designer.
This article proposes a 14-step roadmap from single agent to self-evolving system, emphasizing that base engineering (models, tools, permissions, context) is the key to determining the quality of loop output, and details practical methods for building an efficient base such as CLAUDE.md, sub-agents, skills, hooks, and state files.
Warp CEO Zach Lloyd proposed a two-loop method for an AI Agent's Skill to self-evolve from user feedback, using GitHub issue auto-triage as an example. The inner loop processes new issues, while the outer loop collects signals and distills rules. The framework oz-for-oss has been open-sourced.
Shared a practical website that collects ready-made Agent Loop templates. The core idea is to let the AI agent automatically complete tasks. It includes common loops like Test Until Green, and is suitable for use with Claude Code and Cursor.
The author reflects that the primary bottleneck in running AI agents is not the model's capability but the human's ability to precisely define what 'good' or 'done' means, drawing parallels to managing people.
Developers are moving from one-shot prompts to an 'agentic loop' model, letting AI autonomously set goals and iteratively execute tasks using tools. Claude Code developer Boris Cherny has abandoned traditional IDEs entirely, now running hundreds of agents to monitor issues and merge PRs—even doing so from his phone.
Sydney Runkle announced support for rubrics in deepagents, allowing agents to define a clear definition of done and loop until the goal is complete.
Kocoro is an open-source macOS agent loop and AI runtime engine with local computer access, MCP support, and named agents. It can be used via CLI, daemon, or Kocoro Desktop GUI.