Build an AI Bookkeeper with Pi Agent Harness

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: Build an AI Bookkeeper with Pi Agent Harness
- URL: https://www.youtube.com/watch?v=Khfiy1lwGPs
- Topic: Agent Architecture
- My current learning frame: Build your own single-domain agent on Pi: scaffold the NPM project, create an agent home with agents.md linking to a brain.md of your domain rules plus placeholder skills, then test that it loads its brain and writes structured data to a memory file.
- Why this matters: New playlist item from ZazenCodes; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:00 / Evidence 1: "Today, I'm showing you how to build an AI agent. And to show you how to do that, I'll be building it from scratch myself, so you can follow along. Now, this won't be a vibe coding tutorial."
- 2:42 / Evidence 2: "coding agent. In this demo, I will be using Claude Code. And you're going to need a clear use case. I'm building a bookkeeping agent. You should build something you need. You should build something else. Like right..."
- 4:18 / Evidence 3: "in seeing this, you can skip ahead when I start using Claude Code. And there's no reason you need to actually understand the following stuff unless you're a developer and you're interested in development with Node and TypeScript."
- 8:04 / Evidence 4: "get a coding agent to do all this stuff nowadays. I just wanted to show you this for fun. Now we're going to start using Claude code. So if you've skipped ahead, this is what we've created and..."
- 10:07 / Evidence 5: "agent. But I want to show you this right away before we build the UI. I'm going to ask Claude, "How can I run Pi from inside of this project?" Uh no, can I run it directly from..."
- 14:59 / Evidence 6: "for stuff and I I sort of settled on this project structure, but this is how I approach work, right? Like many of you do. Let's quickly talk through this. agents.md, we created a sort of test one..."
- 19:15 / Evidence 7: "sort of instruction is what you'll find in the default pi agents.md file. That's why I included that in the prompt because I wanted Claude to to build a proper agents.md which it did. This is looking pretty..."

Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable claims from the video. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A one-page agent harness map with tool boundaries and proof signals.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Intent -> Model -> Harness -> Tools -> Verifier -> Artifact
   - 3 concrete examples that apply the video idea to real agentic work
   - 2 failure modes the video helps prevent
   - a checklist I can use the next time I run Codex or Claude
   - one practical exercise with a clear done signal
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.

Quality bar:
- Make this specific to "Build an AI Bookkeeper with Pi Agent Harness", not a generic Agent Architecture essay.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- If evidence is weak, say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.