I take calls with founders pretty much every week. They pitch me their "revolutionary" AI idea, wanting me to build the MVP. And 9 times out of 10? It’s just a UI on top of an OpenAI API key.
Look, I get it. It’s easy. You can spin up a "Chat with your PDF" app in a weekend. But that’s not a business. That’s a feature. And honestly, it’s a feature that OpenAI is gonna kill in their next update anyway.
I’ve been building software for a while now, and here’s what I tell clients who want to pay me to build this stuff: If your "moat" is a prompt, you’re cooked.
Real value isn't in the LLM. It's in the boring stuff nobody wants to talk about.
When I build actual, money-making MVPs for people, we spend maybe 10% of the time on the AI part. The other 90%? * Handling the messy data imports. * Figuring out what happens when the model hallucinates or times out. * Building the logic that connects the AI to their actual internal tools.
I built an agent system last month. The prompt took me 20 minutes to write. The error handling and state management took two weeks. That’s the work. That’s the moat.
Stop trying to be a prompt engineer and start being a software engineer again. Or at least hire one who knows that "import openai" isn't a business plan.
It’s 2026. The easy money is gone. If you aren't doing the heavy lifting in the backend, you don't have a product.
Am I being too harsh here? Or are you guys seeing this "slop" everywhere too?