Home > Videos

🤖🔗⬆️✅ 12-Factor Agents: Patterns of reliable LLM applications — Dex Horthy, HumanLayer

📝🐒 Human Notes

• 🤏 Focused prompt → ✅ quality response
• 🤖 Strength: 🗣️ natural language → ⚙️ JSON
• 🌍 Context Engineering is everything
  • ✍️ Prompt
  • 🧠 Memory
  • 📚 RAG
  • ⏳ History
  • 🧱 Structured Output
• 🎯 Prefer small, focused agents
• ➡️ Agents should be stateless

🤖 AI Summary

The video discusses several challenges and issues related to building reliable AI agents 🤖 and LLM applications, drawing parallels to traditional software engineering principles.

• ⚠️ Difficulty in achieving high quality with agents [00:45]: It’s challenging to get agents beyond 70-80% functionality 📈, often requiring deep dives 🤿 into call stacks and prompt engineering.
• ⚙️ Over-engineering with agents [01:00]: Not every problem requires an agent 🤖; some can be solved with simpler scripts 📝.
• 🎭 Lack of “agentic” behavior in production agents [01:54]: Many production agents function more like traditional software 💻 than truly “agentic” systems.
• ⏳ Challenges with long context windows [02:27]: The reliability and quality of results 📉 decrease significantly with longer LLM context windows.
• ⚠️ Tool use being “harmful” (in a specific context) [04:26]: The abstraction of “tool use” as a magical interaction ✨ makes it harder; it should be viewed as an LLM outputting JSON 💻 processed by deterministic code.
• 🔁 Naive agent loop limitations [06:21]: Simple agent loops don’t work well for longer workflows 📉 due to context window issues.
• 🐛 Blindly adding errors to context [10:54]: Adding full error messages ⚠️ or stack traces to the context can cause the agent to spin out or get stuck.
• 🤔 Avoiding the choice between tool call and human message [12:25]: Builders often avoid deciding whether an agent’s output should be a tool call or a message 💬 to a human, leading to less effective interactions.
• 🖱️ Users needing to open multiple tabs for agents [12:13]: The current user experience often requires interacting with different agents across various tabs 📑, highlighting a need for agents to be accessible through common communication channels 💬.
• 🏗️ Frameworks abstracting away hard AI parts [15:48]: Current frameworks often hide complex AI aspects of agent building 🧱, when they should instead handle other hard parts, allowing developers to focus on critical AI elements.

📚 Book Recommendations

• 🤖⚙️🔁 Designing Machine Learning Systems: An Iterative Process for Production-Ready Applications by Chip Huyen: This book covers the entire lifecycle 🔄 of machine learning systems, including deployment, monitoring, and maintenance, which is relevant to the challenges of building reliable AI agents.
• 📚 💻⚙️ Software Engineering at Google: Lessons Learned from Programming Over Time by Titus Winters, Tom Manshreck, and Hyrum Wright: Given the video’s parallels to traditional software engineering principles 💻, this book offers insights into building robust 💪 and scalable software, which can be applied to AI development.
• 📖 “Building Intelligent Agents: A Tutorial” by Michael Wooldridge: For a more theoretical understanding of intelligent agents 🤖, this book provides a good foundation 🏛️.
• 🗣️💻 Natural Language Processing with Transformers by Lewis Tunstall, Leandro von Werra, and Thomas Wolf: To delve deeper into LLMs and their applications 🤖, this book provides practical guidance 🧭 on working with transformer models.

🐦 Tweet

🤖🔗⬆️✅ 12-Factor Agents: Patterns of reliable LLM applications - @dexhorthy, @humanlayer_dev

🤖 AI Agents | 🛠️ Tool Use | ⚙️ Software Engineering | 💬 Human Interaction | 🐛 Error Handling | ⏱️ Context Window | 📝 Prompt Engineering@aiDotEngineerhttps://t.co/SFlvXWdD4Y

— Bryan Grounds (@bagrounds) July 12, 2025