Build better products. Hire smarter engineers. Execute flawlessly.

But what I’ve been observing in the market is equal parts fascinating and deeply uncomfortable. Engineering excellence is becoming invisible to users.

Take the browser-based code editor space. StackBlitz spent seven years building WebContainers—an entire browser-based operating system that runs Node.js without the need for a server. Replit built its own containerization system, collaborative editing engine, and even its own programming language. That’s years of genuine computer science innovation.

Then Lovable shows up six months ago with what appears to be Claude Sonnet, a text editor, and unlimited marketing dollars.

Today? They’re neck and neck in user perception.

Because users attribute all the improvement to the product, not the underlying model, they don’t know it’s Claude. But perhaps most important, they don’t care.

An Uncomfortable Truth About AI Product Development

What this means is that while AI promises to improve product development, there’s not a lot of actual innovation happening.  It’s just a lot of existing tools in a shiny wrapper.  

Lovable isn’t transforming coding. It’s Claude with prettier buttons. Most AI automation tools aren’t building the future. They’re Zapier with ChatGPT steps.

The technology already works right now. The innovation is permissions and packaging. And this seems to be creating two paths for product leaders:

Path 1: Wait until the LLM actually works perfectly, then scramble to build your ChatGPT wrapper alongside everyone else who just realized the same thing.

Path 2: Start now while the tech is still imperfect. Claim the territory. Build the brand. Survive until the models catch up to your vision. 

In categories where LLMs are rapidly improving, R&D spend on marginal technical improvements might be a vanity metric. Every six months, another 50% model improvement erases months of engineering advantages.

While Replit makes its product 5% better, Lovable ensures 5% more people know they exist. When the next model drops and erases technical differences, distribution becomes everything.

But the companies that lean into this dynamic aren’t necessarily building better products—they’re positioning themselves to be the default choice when everyone realizes the underlying tech has caught up.

And that’s fine, but while they’re busy building awareness, what they’re not building is stickiness.  

The New Reality

We’re entering an era where being early and visible matters more than being technically superior, at least in AI-powered categories. This is part of why we’re in an AI bubble.  

I’ve spent my career believing that great user experiences and solid engineering create sustainable competitive advantages. That’s still true in the long run. 

But in AI-powered categories, the long run is measured in model generations, not years.

This isn’t a temporary market anomaly. It’s a fundamental shift in how product competition works when the underlying technology improves in ~50% chunks every six months.

But simply building more products more quickly is not a sustainable answer when the bubble bursts. These products may allow a company to capture the market, but they’re not focused on retaining the customers who brought them there.  That might not matter much now because everyone’s just focused on grabbing new… but it will. 

The companies that recognize this dynamic early—the ones building products based on retention and operational metrics—are the ones positioning themselves to win when the dust settles. Of course, speed matters more than ever now, but you must still bring quality to the equation to make it in the long run. 

Otherwise, when all you rely on is marketing metrics in your product development, all you’re left with is an empty wrapper that’s been tossed in the trash and an endlessly responsive backlog draining cash and time.