Software Engineers Are Not Dead

“Software engineers are dead.”

I keep reading this from people who have never built a production system… at least not at scale.

Three groups keep saying it:

1. AI researchers

They talk about:

• Transformers
• Papers
• Model benchmarks

Important work. But most of them don’t build production systems.

2. Prompt engineers

They share things like: “10 prompts to make ChatGPT 10x better.” Prompting matters. But prompting is not the hard part of AI systems.

3. Indie hackers

They ship fast. Respect for that. But many have never had to deal with:

• Distributed systems
• Reliability under load
• Identity and auth
• Security boundaries
• Observability

The truth is simple. Real AI products aren’t demos anymore. They’re systems.

Now here’s the irony: AI products need more engineering than traditional SaaS.

Real AI systems require things like:

• AI microservices
• LLM reliability layers
• Latency control
• AI security
• Auth for AI agents
• Prompt evaluation pipelines
• Observability for model behavior

These are hard backend problems. Most people talking loudest about AI can’t build those layers. They don’t want to. Or they don’t know how.

And the companies rushing to sell AI? They will need all of it.

So when I hear: “Software engineers are dead.” I think the opposite.

The next decade will require the best engineers we've ever had. Just a different kind.

AI systems engineers. People who build AI systems, not AI demos.

For more on building with AI in hiring and screening, see our blog and screening tools.