The Ai Product Engineer

There's a role emerging in tech that doesn't have a clean title yet. Let me describe it and see if it sounds familiar:

You sit in product meetings and talk about user problems, business metrics, and roadmap priorities. Normal PM stuff.

Then you go back to your desk and open a code editor. You prototype the feature you just discussed. You test three different models. You write an eval suite. You ship a working demo before the sprint even starts.

You're not waiting for engineering to "pick it up." You're not writing a 40-page PRD that nobody reads. You're building.

You're an AI Product Engineer.

Why This Role Exists Now

Three things converged:

1. AI tools made building accessible. Cursor, Claude, ChatGPT—you can go from idea to working prototype in hours. The technical barrier dropped by 10x.

2. AI products require product-engineering fusion. You can't spec an AI feature the way you spec a CRUD app. The quality is probabilistic. The costs are variable. The failure modes are novel. You need someone who understands BOTH the product implications and the technical architecture.

3. Companies can't afford the handoff tax. The old model—PM writes spec → engineer builds → PM tests → back and forth—is too slow for AI. The iteration cycle needs to be hours, not sprints.

What AI Product Engineers Actually Do

Monday: Prototype a new AI feature using Claude's API. Test it with 50 sample inputs. Measure accuracy.

Tuesday: Present the working prototype to stakeholders. Not a slide deck—a live demo. "Here's what it does. Here's where it fails. Here's what it would cost at scale."

Wednesday: Design the eval framework. Write automated tests. Set up the LLM-as-judge pipeline. Define "ship quality" with actual numbers.

Thursday: Work with the engineering team to production-ize the prototype. You're not handing off a spec—you're handing off working code with test coverage.

Friday: Monitor production metrics. Spot a quality regression. Debug it. It was a prompt that worked great with GPT-4 but broke with the latest model update. Fix and deploy.

This isn't theoretical. This is what the best AI builders do every week.

The $200K+ Premium

Companies are desperate for this person. Here's why:

• A traditional PM can define what to build but can't validate feasibility quickly. They write specs based on assumptions about AI capabilities.
• A traditional engineer can build what's asked but may not optimize for the right user outcomes. They over-engineer or miss the product insight.
• An AI Product Engineer does both. Faster. With better judgment about what's worth building.

The salary data backs this up. AI Product Engineers (even when they carry a "PM" or "Engineer" title) command $250K-$500K+ at top companies. The supply is tiny because this combination of skills barely existed two years ago.

How to Become One

You don't need a CS degree. You don't need to quit your PM job. You need to:

1. Start prototyping. This weekend. Pick a feature idea, open Cursor or Replit, and build a working version using an LLM API. It doesn't have to be good. It has to exist.

2. Learn evals. This is the #1 skill that separates AI Product Engineers from everyone else. If you can design an eval framework, you can ship AI features with confidence.

3. Understand model tradeoffs. Not just "GPT-4 vs Claude"—understand latency, cost, context windows, fine-tuning, and when each matters.

4. Ship something. Deploy a real AI feature. Even a small one. Even an internal tool. The gap between "I understand AI" and "I've shipped AI" is enormous.

5. Build in public. Share what you learn. Write about your failures. The AI Product Engineer community is forming right now—be part of it.

The PMtheBuilder Thesis

This newsletter has always been about PMs who build. The "AI Product Engineer" framing just makes it explicit.

If you're reading this, you're probably already one—or becoming one. The title on your business card might say "Product Manager" or "Technical PM" or "AI PM." Doesn't matter. What matters is what you ship.

PM the Builder is for people who ship AI products. Not people who strategize about them.

Welcome to the build.

---

Next issue: How to design evals that actually matter (the AI Product Engineer's most important skill)