Agent Systems: design, test and govern agents at scale

Copy this, open Claude, paste it, and Claude will walk you through the lesson step by step.

For a grown-up: Claude is for adults (18+). Copy this, paste it into Claude, and use the answer to explain the lesson to your child.

I'm learning "Agent Systems: design, test and govern agents at scale" on The Future Market. Lesson 1: "One agent or many? Choose the simplest thing that works".

What the lesson teaches:
- Anthropic's article "How we built our multi-agent research system" describes an orchestrator-worker pattern: a lead agent plans the work and sends subagents to do parts of it in parallel. Subagents search, check results and report back to the lead.
- In that system, the multi-agent version beat a single agent on Anthropic's internal research evaluation by 90.2 percent. The article also says that agents use about 4 times the tokens of a chat, and multi-agent systems about 15 times the tokens of a chat. (These are the figures from that article, for that system. Your numbers will differ. Measure your own.)
- The article's conclusion: multi-agent systems pay off for high-value tasks that split into many independent parts. They are a poor fit when the work has many dependencies between steps, such as many coding tasks.
- From Anthropic's "Building effective agents" guide: start with the simplest solution and add complexity only when it clearly improves results. Many jobs need only a single call, or a fixed workflow, and not an agent at all.
- The article gives effort rules of thumb: a simple fact-finding job needs 1 agent and about 3 to 10 tool calls; a direct comparison needs 2 to 4 subagents with about 10 to 15 calls each; complex research can use more than 10 subagents. Tell the lead agent how much effort a question deserves, or it may over-spend.
- Claude Opus 5 delegates to subagents more readily than earlier models, according to the Agent SDK docs. That makes depth, concurrency and spend limits (Lesson 6) more important.
- Question - If yes - If no

The exercise I'm working on:
Take a job from your own work or study (for example "prepare a weekly report from five sources"). Fill in the decision table, build the simplest version, and write the one-paragraph decision. Do not build the multi-agent version yet.

I'm stuck. Please:
1. Explain the main idea in plain, simple words with one everyday example.
2. Walk me through the exercise one small step at a time. Ask me what I've done so far before giving the next step.
3. Don't just give me the final answer. Help me do it myself, and check my work when I share it.

I'm a parent helping my child (about 8 years old) with a lesson called "One agent or many? Choose the simplest thing that works" from the course "Agent Systems: design, test and govern agents at scale".

Here's what the lesson says:
A small bakery needs one baker. A huge wedding order may need a head chef who hands out jobs: one person on cakes, one on bread, one on icing. But more cooks cost more money, and they need to talk to each other. So before hiring more cooks, ask: is one good cook enough?

The activity:
Take a job from your own work or study (for example "prepare a weekly report from five sources"). Fill in the decision table, build the simplest version, and write the one-paragraph decision. Do not build the multi-agent version yet.

Please explain this so an 8-year-old understands: short sentences, a fun everyday example, and no hard words.
Then give me 3 simple questions I can ask my child to check they understood, and a way to do the activity together, one tiny step at a time.