✨ the SPARK 253 ~ I Built My AI Quiz in 2 Weeks (Part III: The Messy Middle)
What I'd Tell Someone Starting This (And Why It Matters)
*This is the 3rd & final part of rebuilding my AI quiz from scratch. Read Part 1 and Part 2 here.
Last week, I told you about the visual decisions and UX thinking that went into the quiz. This week? We’re talking about the part that actually took the longest—the part nobody sees.
The quiz is live at FindYourAIpath.com.
And honestly? The prettiest parts were the easiest. It was everything you can’t see that gave me a bit of a headache and an “I’ll be back tomorrow to finish this” (with Claude).
What “Personalization at Scale” Actually Looks Like
Someone takes the quiz, opts in, and receives a customized 5-day email sequence based on their results. Not generic, not “insert name here.” Actually personalized to their specific combination of Primary Path, Secondary Strength, and Skill Level.
Here’s the results page (cut in half, it’s all on one page):
Nine paths, three skill levels, multiple combinations. All flows automatically into my email system with the right data in the right format.
I used Claude to work through the logic, plan the data structure, and troubleshoot the conditional statements. I knew what I wanted people to experience, but AI helped me figure out how to make it happen step by step.
This is where AI shines, not replacing your vision, but helping you execute it.
Here’s where it gets interesting (er, challenging? lol).
The Case Sensitivity Nightmare
Mocha (the vibe coding software I use) sent over one path name in lowercase: “content integrator” instead of “Content Integrator.”
My email platform expected proper capitalization for the conditional logic.
Result? Everyone with that specific result would get generic fallback text instead of personalized content.
The fix took 30 seconds. Finding it took a tad longer, comparing data fields, and questioning my sanity. I caught it because I was diligent in testing it (i.e., not just trusting Claude’s output).
Universal Lesson: When systems talk to each other, confirm they’re actually speaking the same language. Provide feedback on what you think is happening at each step and verify that it’s what’s supposed to be happening. The slightest misalignment, a capitalized letter, a space, or a comma, can break everything silently.
The 12-Pixel Font Disaster
I spent hours crafting personalized email sequences with conditional logic.
Then I sent myself a test email. The personalized sections (custom fields) defaulted to 12px font size… the rest of the email was 16px.
It looked like someone had copied and pasted from a 1995 website (seriously… who reads 12pt font?!?! #gettingold).
I tried both versions of the snippet editor (snippets pull custom subscriber results). Didn’t work. Finally, I put the code directly in the email body and forced the font size.
Universal Lesson: Automated email systems often use default formatting that looks terrible. Test your actual emails; there’s usually a workaround.
When Simple Beats Sophisticated
Here’s what happened with the integration:
Tried connecting Mocha to Kit via API (the “proper” technical approach)
Didn’t work
Went back to basics: embedded Kit’s direct opt-in form on the results page
Voila! Easy peasy
Then there was the personalization code. Kit uses Liquid to display different content to different people based on their quiz results. I needed the emails to display different paragraphs based on a person’s specific combination of results.
The way it was written wasn’t straightforward. The documentation wasn’t clear for my use case.
Universal Lesson: Sometimes the “less sophisticated” solution is the one that actually works. Start simple, get it working, and add complexity later if you actually need it.
What I’d Tell Someone Starting This (And Why It Matters)
If you’re thinking about building a personalized tool—a quiz, an assessment, a calculator, whatever—here’s what I’d recommend:
Get crystal clear before you build anything. What’s the value to the user? What do they actually get out of this? Then map out the entire structure: how the logic works, what data you need, how it flows from one system to another. This is where AI can help you think through everything in natural language before you build anything (there’s a reason my WisprFlow dictation count is so high, lol).
Test as you build, not after. Test each piece the moment you create it. Don’t wait until everything’s done. Test combinations you think won’t happen. Test with real accounts, not just preview mode.
Platform quirks are inevitable. Every tool has weird limitations. Accept this early, and you’ll save yourself the frustration of thinking you’re doing something wrong.
Simple beats sophisticated. If the basic integration works, use it. You can always upgrade later when you have proof of concept and real user data.
This isn’t just about building one quiz. It’s about understanding what goes into creating tools that actually work—not just tools that look good in a demo (or are visually appealing… vibe coding apps are misleading this way).
The difference between “it works” and “it works well” is in the invisible infrastructure. And this is where most people give up. They build the front end, realize the backend is more complicated than expected, and either ship something half-functional or abandon it.
When you hit these challenges, they’re not a sign you’re doing it wrong. They’re just part of building things that actually work.
Test It Yourself
The quiz is live: FindYourAIPath.com
Take it. See the personalization in action. You’ll get your results immediately, followed by a 5-day email sequence customized to your specific results (you won’t be on my list twice 😉).
And if something breaks? Tell me, because that’s the other thing about building in public. You get honest feedback from real users, not just your own assumptions.
But here’s what I want to be really clear about: getting it live is just the beginning.
Over the next couple of months, I’ll be watching the data. Where are people dropping off in the quiz? Which questions might be confusing? Which email in the sequence gets the lowest engagement? Are certain paths converting better than others?
This is how you create something that works for both your subscribers and your business—you launch it, measure it, adjust it, and keep iterating based on real user behavior.
Version 1.0 is live. Version 2.0 will be informed by actual data, not just my assumptions.
I’ll also run paid traffic to the quiz in Q1 of 2026, so we’ll see how it performs.
SPARK Spotlight 🔥
The quiz, of course.
I’m still editing and tweaking some of the text in the follow-up sequence (or maybe by the time you get this, it will be done 😉), but if you’d like to take the “Find Your AI Path” quiz, here you go!
There are a few other technical things I’m still tweaking, but I won’t bore you with that. Any feedback is appreciated (clarity, technical issues, whatever).
A Little Brainpower 🧠
Yet another reason to love Nano Banana Pro. “I tried using Nano Banana Pro to create complex infographics,” by Karen Spinner.
Next year will be interesting to say the least. “Where Substack is Headed in 2026,” by Mack Collier.
This is the power of AI. “This Million-Dollar Founder Has No Marketing Team. Just 40 AI Agents.”
Tool Time 🛠️
Tasks: Organize your tasks with AI-powered insights. Sync across devices and never miss a deadline. Try SnapToDo free here.
Visuals: Transforms content into stunning visuals, such as infographics, from existing text, files, videos, and web searches. Try ViziVibes free here.
Documentation: an AI documentation tool and platform that helps create and publish up-to-date documentation and knowledge bases. Try Documentation.ai free here.
Calendar: Connect multiple calendars & meeting tools & go-live in 60 seconds. Try Luncal.ai free here.
Haha… if only this weren’t true 😂
We’re almost halfway through December, and if you’re feeling like you’re running on fumes… ditto. 🤪
The holidays are beautiful and exhausting at the same time. So if you need to coast into the new year, coast.
For me? The quiz is built. The infrastructure is solid. The data will come.
Right now? I’m taking a breath and enjoying the “fa-la-la-la-la” phase while it lasts. January will bring its own energy (and probably some melting snowmen), but we’ll deal with that when we get there.
For now, we’re almost there.
That’s enough.
With lots of coffee & kindness,
Kim
Very cool, Kim! I did the quiz ✨✨ Loved it and very accurate for where I am right now!
This line here: "The slightest misalignment, a capitalized letter, a space, or a comma, can break everything silently."
Only someone actually building AI tools and automations will understand this. Can be so maddening at times!