Why This Matters
Rise of AI Agents. From the people who brought you LangChain and LangGraph — Deep Agents represents the next evolution in autonomous AI systems. In a blog post for Medium readers, the important thread is not only what happened in the training, but why it changes how a team should think or work. Keep the framing conversational, practical, and personal.
Core Ideas
Ability of AI doubles every few months, leading to rapid advancements in capabilities and the emergence of autonomous AI agents. What they do. Use more tools. Use planning. Task Delegation(Sub agents). Use file to offload context. Use prompt engineering for Prompt Engineering. Planning. Use a todo list todo.md. Planning for user approval (ask the user to approve the plan or ask follow up questions before proceeding). Re read plan at select points and time to make sure that the agents report. Iteravely update the todo.md. Task Delegation. Use sub agents to delegate tasks. watch for: Conflicting decisions. File System. use filesystem to offload contect and data. Offload the raw tool observations though a summary. Uses a file to store the raw observations so that it can look at it if it needs to. Keep the framing conversational, practical, and personal.
How It Works in Practice
Prompt Engineering. Understanding prompts can help make AI agents better. Steer Agent. Create Agent. Define the agent’s goals and capabilities. Set up the initial context and environment for the agent. Specify the tools and resources the agent can use. Implement the agent’s planning and decision-making processes. Test the agent in a controlled scenario before deploying it for real tasks. TODO Lists. Plan. State. todo. files. Keep the framing conversational, practical, and personal.
Takeaways
Make a todo. We need to give our agent a .md on how to create todos. WHen to use. strudture. Best practices. Progress update. Parematers. returns. Explanation of how to use files to offload context and data for AI agents. SUbsgent. Just another tool call, similar to how other tools are used by the agent, allowing the agent to delegate tasks to sub agents as needed. The practical value is in turning these notes into decisions, checks, and habits that can be reused after the training is over. Keep the framing conversational, practical, and personal.