Can a Marketer Vibe-Code a Working App? 6 Lessons From My First Build

Can a Marketer Vibe-Code a Working App? 6 Lessons From My First Build

I spent 10 years in digital marketing and SEO and built every site with WordPress.

This year, I decided to try to build a finance dashboard to track corporate insider and politician stock trades.

WordPress didn't seem right for this, so I decided to try vibe-coding.

I built the site, called Insider Hawk, with Lovable.

It's a single-page application (SPA) built with React.

Here are the takeaways and what I plan to do next with the project:

1. Cursor, VS Code, or Lovable?

I have a friend who is an experienced programmer who told me to try vibe-coding with Cursor or VS Code.

I found these programs to be a bit cumbersome to start with.

Dealing with terminal commands and server setups felt slow and disrupted the momentum.

So I tried Lovable.

I think for a first version, Lovable was fine, and I like that it syncs to GitHub, meaning you own your code and database and can move them off the platform if needed.

Side note: More recently, friends have been recommending Claude Code desktop app, so I do want to try that for future projects.

2. Two steps forward, one step back

As a non-programmer, building the initial prototype of my site took some patience and trial-and-error.

Sometimes, I'd build a feature that appeared to work, but on the backend, it wasn't functioning at all how I expected or wanted.

For example, for my main homepage that would show the latest corporate insider trades, I assumed we'd collect data in a database and then the front-end would be generated using that data.

Yet because I didn't specify that architecture at the start, the webpage was fetching completely fresh data from the APIs for every single visitor.

If I remember correctly, a lot of calculation and page generation was being dumped directly into the user's browser, too.

So I had to discover that and fix that, and start leaning on our database (Supabase is what Lovable uses by default)

3. "You don't know what you don't know"

My programmer friend had to tell me certain things that I would never have known without him.

For example: Proper data ingestion and data sanitization from our APIs. Not just letting any/all data hit our database.

I had zero concept of this before my friend explained it to me.

So if you're a marketer vibe-coding your first project, try to ask a programmer friend for tips as you go, or find a few good online communities/forums to join.

4. The AI makes mistakes

Lovable, which uses a mix of various AI models including Claude and others, does make mistakes.

I specifically asked it to grade politicians' trading track record based on 365 days, and it even wrote the little tool-tip description that said it was based on 365 days.

Yet I discovered the actual code/logic was only looking at 30 days, not 365.

I noticed because I saw Nancy Pelosi, who usually performs incredibly well, showed a 1-star rating on the politician trades view.

So I asked the AI how this was possible and it informed me that this was indeed her performance over the past 30 days. It didn't recognize that we had never discussed using 30 days and that this was wrong.

We had only discussed and agreed on 365 days.

5. On-page SEO is not good out-of-the-box

Lovable recently rolled out an SEO tool to audit/fix issues. They have a similar tool for website security.

This is a good start, but I found that I still had to fix some on-page SEO issues myself.

It made me realize how good I had it with my previous content sites built on WordPress. WordPress is very solid out-of-the-box for SEO, and a React site built with Lovable is going to take a bit more setup/fixing.

6. Conclusion: Worth it?

I've got many other ideas of single-page web apps, and I think it will take me 5-10x less time to build the next one.

I've learned to send better prompts, think about potential issues before they arise, and fix mistakes faster.

I'm looking forward to building out any future ideas/projects.

It was absolutely worth learning to vibe-code.

Being able to turn an idea into a functioning website is incredible. Programmers have had this "superpower" for decades. Now I can take advantage of this as well.

For a "v2" of Insider Hawk, I might hire a developer to ensure it's well-optimized, efficient, and scalable.

Also, I'd hire a pro if I ever start having user logins/accounts, taking payments, etc.

But since my website currently does not have logins, accounts, or payments, I wasn't too worried about vibe-coding a "v1".

If you're curious how the site looks and functions now, you can check it out at InsiderHawk.com