Summary: The capstone

For education only. This teaches multi-agent architecture using stock analysis as the example; it is not investment, financial, or trading advice.

The capstone is where the whole team you studied runs at once, and where you can run it yourself. It uses the open-source TradingAgents framework (anchored to one frozen snapshot, marked 7e9e7b8), the same code you read across the track. This summary is the scan-in-five-minutes version of the full lesson.

Core ideas

• The pipeline runs the path you studied. The four analysts gather, the bull and the bear debate, the research manager rules, the trader plans, the three risk seats stress-test, and the portfolio manager makes the final call. A live status board shows each agent flip from waiting to done, grouped by team.
• Seven reports, one per stage. The run writes plain reports to a folder: the four analyst reports (lesson 2), the investment plan (lesson 3’s judge), the trader’s plan (lesson 4), and the final trade decision (lesson 5’s gate). Each file is an agent you already met.
• The two judges are visible in the output. The research manager turns the debate into a ruling; the portfolio manager (the deep-model gate) reconciles the three risk voices into a final written recommendation on the five-point scale. It commits, it does not average.
• You can run it, safely. It is the same open-source code you studied, pinned to one snapshot, run from a single command after a short setup. It is a simulation: it writes a research opinion and a set of reports, and it places no trades. You supply your own AI provider key.

What changes for you

The architecture is the lasting takeaway, not the domain. You have now watched a complete pattern run end to end: focused specialists, tool-using data gathering, forced disagreement, deciding separated from doing, a well-resourced gate with veto power, one shared source of truth, and learning from real outcomes, with the most capable model spent only at the points of decision. That pattern travels to research, operations, support, drafting, and review. The next time you build something real with AI, you have a studied example of what good structure looks like, and the confidence to ask where the specialists, the disagreement, and the gate are.