You finished the curriculum
What's next?
Reading is one thing. Shipping is the other. Pick a project, build it, and bring it back to the community. Below are four starting points; pick whichever fits the kind of agent you want to live with.
Build a research assistant
An agent that takes a topic, plans a query strategy, retrieves from multiple sources, and writes a synthesis with citations. Use what you learned in Track 1 (RAG) and Track 2 (orchestration topologies).
Replace one of your tools with an agent
Pick something you do manually every week. Spec it as a goal, list the tools the agent needs, write the loop. Ship it as a CLI you'd actually use.
Publish an MCP server
Expose a service you depend on (calendar, notes, your file tree) as an MCP server. Wire it into Claude Code or another MCP host. Track 3 has the patterns.
Run an agent in production
Deploy one of your sandbox exercises behind a queue, add structured logging and an eval, and run it on real input for a week. Track 4 is the playbook.
Show what you built
Post your project to the community subreddit. Include what it does, what surprised you, and what you would do differently next time. Other learners read these and skip mistakes you already made.