Code Less, Build More (or get left behind)
The edge now belongs to those who can build with AI, not just code around it.
Who Builds With AI?
I interview engineers almost every week for clients and for startups I work with (for).
Over the past year, something's shifted.
I’ve found myself rewriting interview questions, adding practical AI scenarios, and evaluating an entirely new kind of engineering instinct. The resumes haven’t changed much, but the skillset has. There’s a new kind of developer in town. Not just a coder, not quite a researcher, but something in between.
They don’t train models from scratch. They assemble, integrate, and orchestrate. They ship full apps in days, not weeks. They are GenAI Application Engineers, and they're quickly becoming indispensable. (Not sure I like that term, but let’s use that for now) It seems good enough for Andrew Ng)
The shift isn’t subtle. AI-native applications are the new normal. Tools like GPT, Claude, and Midjourney aren't just novelties anymore—they’re product infrastructure. Someone has to figure out how to build with them, and fast. That someone is this new breed of engineer.
The Problem
Traditional engineering roles aren’t built for this pace.
AI capabilities change monthly, sometimes weekly
Product needs evolve faster than backlogs can keep up
Most engineers are either too "deep tech" or too "vanilla full-stack"
We’re in an awkward adolescence where GenAI is powerful, but most companies still don’t know how to use it. Teams try to bolt LLMs onto existing products and wonder why it feels janky.
That’s where the GenAI Application Engineer thrives.
Insight From the Field
The stack is weird. The job is real.
These engineers don't think in terms of REST APIs and React components alone. They think in:
Prompt engineering and Retrieval-Augmented Generation
Agentic frameworks like LangChain, CrewAI, AutoGen
Vector databases (Pinecone, Qdrant) and LLM APIs
Evals, guardrails, fine-tuning, and orchestration logic
They also use AI to build AI. Tools like Claude Code, Cursor and Zencoder act as pair-programmers that can complete, debug, and test entire modules. The goal isn’t just to code—it’s to ship faster and smarter.
The best GenAI Engineers come from full-stack backgrounds, but they’ve layered on AI-native skills. They’re not reinventing transformers. They’re composing them like APIs.
Lessons to Take Away
This is application-first engineering It’s not about building the model. It’s about building with the model.
Prompting is a skill, not a job Prompt engineering matters, but it's just one tool in a broad kit.
Good product sense is a force multiplier The best GenAI Engineers are also great at scoping, prototyping, and iterating. They build what users want, fast.
The role is still forming Titles vary. Tooling is chaotic. Standards are fuzzy. But make no mistake—this is the new full-stack.
Hiring for this is hard It’s not easy to screen for skills that didn’t exist two years ago. Look for people who learn fast and ship fast.
Closing Thoughts
If Agent Bosses are the new “managers”, GenAI Application Engineers are their architects. They are not a fad. They’re the backbone of every AI-native product being built today.
The stack may stabilize. Titles might get formalized. But the demand for engineers who can build with GenAI is only going up.
Every product team needs at least one. Every founder should be one.
And every engineer should ask:
What can I build if AI is part of the stack from day one?