Capstone: a Gem and a notebook over the same sources

What this capstone proves

Everything in this level taught a feature. This capstone asks you to hold two of them side by side and show judgement. A Gem and a NotebookLM notebook can both answer from your files, so the interesting question isn't "can they?", it's "which one, when, and why?" Answering that with evidence is what separates someone who uses Gemini from someone who chooses the right part of it.

The heart of the exercise is the comparison. Anyone can build a Gem and build a notebook; the champion move is to put the same real questions to both, notice honestly where each is stronger and weaker, and turn that into a one-line rule you'd actually follow under pressure. A capstone that ends "they're both quite good" has failed. One that ends "for a recurring drafting job in Fernway's voice I use the Gem; for a traceable answer I'll quote in a meeting I use the notebook" has succeeded.

The shared source set

Both tools will sit over the same Fernway paperwork, so the only variable is the tool. Use this set: the project brief, the meeting notes, the email thread, the remote-working policy and the sales spreadsheet. It's a realistic mix (a plan, some rough notes, a live discussion, a policy and a messy data file) which is exactly where the two tools start to behave differently.

The project brief

The comparison write-up template

Copy this and fill it in. Keep it to a page. A champion communicates a judgement briefly.

Gem vs notebook: same sources, side by side

• Source set used: the five Fernway files (or your equivalents)
• The Gem in one line: its role, format and key limit
• The notebook in one line: its sources and what you asked of it
• Question 1 (drafting): which tool won, and why: voice, format, editability
• Question 2 (factual lookup): which tool won, and why: traceability, citations, gaps flagged
• Question 3 (data + judgement): which tool won, and how each handled the messy sheet and typos
• Where the Gem was stronger: e.g. reusable, on-brief drafting in Fernway's voice, one-click
• Where the notebook was stronger: e.g. clickable citations, easier to verify, Audio Overview, exploration
• Anything either invented or misread: be specific, this is the honesty test
• My one-line rule: "Reach for the Gem when… ; reach for the notebook when…"

Suggested approach

A useful way to think about the two, before you test them: a Gem is a doer you configure once. You bake in a role, a house style and a format, and it produces on-brief output (a draft, a summary in your shape) with a single click, again and again. Its strength is reuse and voice: the recurring job done consistently. A notebook is an interrogator pinned to your sources. Its strength is traceability and exploration: every claim links to the passage it came from, and an Audio Overview lets you get across the whole set hands-free. Roughly: the Gem is better when you'll do the same shaped job repeatedly and want it in your voice; the notebook is better when you need to explore a fixed set of documents and prove where each answer comes from.

Test that hypothesis rather than assuming it. Put the same questions to both and watch what actually happens. You'll likely find the Gem drafts a more usable reply in Fernway's tone, while the notebook makes the factual lookup easier to trust because you can click every citation. On the messy sales question, watch both carefully. Neither reading a mistyped "Sotuh" as "South" is a tool failing you should catch and note, not smooth over. Let the evidence, not the theory, write your final rule.

Above all, keep yourself as the check. The point of building both isn't to find the one you can trust blindly. It's to understand each well enough to know what you still have to verify in each. That judgement is the deliverable.

Self-assessment rubric

Score yourself honestly against each level, and be able to point at evidence for your score.

• Basic. You built a working Gem and a working notebook over the same sources and can describe what each did. Your comparison is impressionistic ("the notebook felt more trustworthy") rather than grounded in specific test results, and you didn't really probe where either failed.
• Good. You built both properly (the Gem with layered instructions and a real limit, the notebook with the full source set and an Audio Overview), ran the same three questions through both, and recorded concrete differences. Your write-up names where each was stronger and ends with a clear one-line rule you'd actually follow.
• Excellent. All of "good", plus: at least one of your test questions deliberately stressed the tools (the messy sales data, a gap the files don't cover), you caught and documented a real misread or invention in one of them, you tied each tool's strength to why it behaves that way (instruction-shaped reuse vs source-locked traceability), and your final rule is specific enough that a colleague could apply it without you. Someone reading your one-pager would know exactly which tool to pick for their own task.

The gap between "good" and "excellent" is almost entirely about stressing the tools and being honest about what broke, not about building anything fancier.

Evidence note

Common mistakes

• Building both but never comparing them properly. Two tools sitting side by side isn't the deliverable. The judgement is. Without the same questions run through both, you have two demos and no comparison.
• Different questions for each tool. If you ask the Gem one thing and the notebook another, you can't tell whether a difference is the tool or the question. Keep the test questions identical.
• Only asking questions that flatter them. A capstone that avoids the messy sales data and the gaps in the files never finds the failure modes, which is where the real learning, and the "excellent" grade, lives. Stress them on purpose.
• Over-trusting whichever one you preferred. Deciding the notebook is trustworthy because it has citations, or the Gem is right because it's anchored, is the exact trap. Citations can point at a misread passage; an anchored Gem can still fabricate. Verify a load-bearing answer from either, and record where you had to.
• Writing a vague rule. "Use whichever suits the task" helps no one. The valuable output is a specific, followable rule: recurring on-brief drafting → Gem; traceable, quotable answers over a fixed set → notebook.

Keeping current

Both tools are developing quickly. Gems' knowledge limits and sharing, and NotebookLM's sources, Audio Overviews and availability all change. Your method here (build both, test with identical questions, compare honestly, keep yourself as the check) is durable and will outlast any specific feature. When you revisit this, re-check each tool against Google's official pages: Use Gems in Gemini Apps and NotebookLM Help. Accurate as of 13 July 2026.