The Tool Doesn't Make the Designer: A Vibe-Coded eLearning Experiment

May 25, 2026

When I was finishing my master's in Instructional Technology, Training, and Development, I spent a lot of time scrolling job boards and LinkedIn trying to figure out what skills I actually needed to land my first ID role. The answer I kept seeing was the same one over and over: learn Storyline.

It was on almost every job description. It was the centerpiece of every bootcamp ad. It was the answer to every "what tool should I learn first?" post in every ID Facebook group. For a while I genuinely thought it was Storyline or bust — that if you weren't fluent in it, you weren't a real instructional designer.

So when The eLearning Designer's Academy ran their May challenge — build a course using Anthropic's new Claude Design instead of a traditional authoring tool — I saw it as a chance to test something I'd been wondering about. Is Storyline really the only path? Or has the field already started to widen, and the job descriptions just haven't caught up?

This post is what I learned.

The Project

The challenge asked us to take a past design challenge and rebuild it using Claude instead of a traditional tool. I picked the April 2025 challenge — bloodborne pathogens training for a fictional senior care facility called Knotty Pines — because the original brief offered six possible interaction types and I wanted to see how many I could pull off in one course.

I ended up hitting five:

• A PPE selection checklist with a final reveal of a fully-equipped caregiver

• A patient room contamination hunt with invisible hotspots (you have to actually look)

• A handwashing technique walkthrough

• A hand-germ prediction interaction

• A branching exposure-response scenario with consequences

Plus a 6-question knowledge check, a printable certificate, dark mode, captions, narration, replay buttons, keyboard navigation, and full state persistence across reloads.

You can click through it yourself here: https://willowy-liger-53df65.netlify.app/ (works best in Chrome or Edge on a desktop).

The Process

Here's what surprised me first: the process wasn't that different from what I'd learned in grad school.

I wrote the full storyboard before opening Claude. Every screen, every interaction, every narration line, every accessibility note. I gathered the style guide, the logos, real photography for every screen. Then I uploaded all of it and made Claude wait until everything was in hand before writing a single line of code.

The structured input changed the output quality dramatically. The instructional design fundamentals I learned in my master's program — needs analysis, learning objectives, scaffolded interactions, formative assessment — all of that still applied. If anything, they mattered more, because the AI builds exactly what you tell it to build. Skip the storyboard and you'll get freestyle results. Bring the same rigor you'd bring to a Storyline project and you'll get something you can actually defend.

Where I Stayed the Designer

The most important part of this project is what I decided, not what Claude decided.

Early on, Claude built the contamination-hunt screen with visible hotspots — little pulsing dots showing exactly where to click. I killed it. The whole point of a "find the risks" interaction is that the learner has to look. Visible hotspots reduce it to a click-the-marker exercise.

On the handwashing screen, Claude initially built a 20-second hold-to-scrub interaction. I realized it was just cycling still images — not actually a video — and a fake timer doesn't teach anyone anything. I cut it and replaced it with a clean visual reference of each step.

I insisted on photorealism images over illustrations — and because I am not in the healthcare field as a senior care nurse, I asked ChatGPT to help me create photorealistic images for the course (this is where ChatGPT outshines Claude). I added a name input on the first screen so the certificate wouldn't be generic. I pushed for warmer, more human narration over the default robotic text-to-speech voice.

None of those are technical decisions. They're instructional design decisions. And Claude was happy to build whatever I told it to build — which is exactly the point.

Where Claude Took Over

The implementation. Layout patterns, persistence architecture, the timing logic for chained narration across multiple cards, every bug fix along the way. I didn't have to know how localStorage works. I just had to know I wanted progress to save.

This is where I'll say something I wish someone had told me earlier in my master's program: I think the conversation about "what tool should new IDs learn" is about to broaden a lot.

Storyline is still the right tool for a lot of work, and the people who know it deeply have built a real craft I respect. I'm planning to keep learning it. But the assumption that it's the only path into this field is starting to feel out of date — and I think the job descriptions are going to catch up faster than people expect.

Where It Broke Down

This wasn't all magic. A few honest tradeoffs:

I became QA. In Storyline, the tool handles state. With Claude, every regression was something I had to catch and report. Caption sync broke twice. The progress bar collapsed at certain screen widths. The replay button worked but lost its captions. Each fix was a round trip.

No visual canvas. I'm a visual learner. Describing a layout in words and then judging it after the fact was the steepest part of the curve. I missed being able to nudge something two pixels left.

Browser text-to-speech isn't production-quality. For a real client deliverable, I'd want recorded VO. The narration works for a prototype; it's not what I'd ship.

What This Means For New IDs

There's a lot of anxiety in the ID community right now about whether AI is going to replace us. I see it in every Facebook group and LinkedIn comment section. As someone who just spent two years and a lot of tuition learning this craft, I had every reason to be nervous walking into this challenge.

I'm not anymore.

Claude didn't make any of the instructional design calls. It couldn't tell me that visible hotspots break a "find the risks" interaction. It couldn't tell me that a fake timer doesn't teach handwashing. It couldn't decide that the certificate needed a real name on it to feel like the learner's own. It built what I asked it to build, and the things I didn't ask for, it didn't catch.

What AI can do is collapse the implementation gap. The accessibility features I'm most proud of in this course — captions, replay, keyboard navigation, reduced motion, full UDL principles — would have taken me significantly longer to build in a traditional tool and sometimes require multiple tools. In Claude, I just asked. That's not a threat to our profession. That's a gift to it. It means new IDs like me get to spend more time on the parts of the work that actually require us: understanding the learner, structuring the experience, making the calls that AI can't.

What I'm Taking Forward

I'm at the very start of my career, and I'm taking three things from this project into whatever comes next.

One: the fundamentals from my master's program weren't optional. They were the thing that made this whole project work.

Two: "Storyline or bust" is a story new IDs tell each other because the job market hasn't caught up to where the tools actually are. I'm not abandoning Storyline — I want to learn it well. But I'm also not going to pretend it's the only thing worth knowing.

Three: if you're another new ID who's been nervous about AI, try it. Bring your storyboard, bring your style guide, bring everything you learned in school. You'll find out pretty quickly that

AI isn't coming for your job. It's coming for your timeline.

🔗 Try the course: https://willowy-liger-53df65.netlify.app/

Best viewed on a desktop in Chrome or Edge. On mobile, turn your phone sideways to collapse the course outline menu.

Connect with me on LinkedIn ❊ View My Work