I discovered something yesterday that gave me pause.

On December 9, LangChain published an article called “ Agent Engineering: A New Discipline .” I’d been developing a framework called Synthesis Engineering since early November. Reading their piece felt like finding notes from a parallel investigation.

The core insight is identical: production AI development needs a discipline, not just tools. The casual approach that Andrej Karpathy memorably called “vibe coding” works for exploration. It fails when systems need to be reliable, maintainable, and safe.

When multiple people independently identify the same problem, it usually means the problem is real.

The convergence

I published my first Synthesis Engineering articles on November 9 — about a month before LangChain’s announcement. The body of work includes twelve articles covering everything from technical implementation to scaling in organizations to case studies from actual projects .

LangChain’s Agent Engineering focuses specifically on agent systems — the deployment lifecycle for AI agents. Synthesis Engineering takes a broader scope, covering all human-AI collaboration in software development, with a distinction between the discipline (Synthesis Engineering) and the hands-on craft (Synthesis Coding).

The frameworks are complementary, not competing. Agent Engineering addresses a specific slice of what Synthesis Engineering covers. Where they overlap, they agree.

Why I released it under CC0

When I developed Synthesis Engineering, I had a choice: keep it proprietary or release it to the world.

I chose CC0 public domain . Anyone can use it. Teach it. Build certifications on it. Incorporate it into commercial offerings. No permission needed. No licensing friction.

This wasn’t altruism. It was strategy.

Ideas spread faster without friction. A framework that requires licensing discussions won’t be adopted as quickly as one that doesn’t. The goal is adoption, not ownership. If Synthesis Engineering becomes how people think about production AI development, that’s more valuable to me than any licensing revenue would have been.

LangChain is welcome to use it however they want. So is anyone else.

What I’m building

I’m planning speaking engagements on Synthesis Engineering and Synthesis Coding. I’m working on a book that will go deeper into the practices. I’m considering a certification program for practitioners.

All of these could be collaborative rather than parallel. There’s room in the market for multiple voices and approaches. The industry benefits from shared vocabulary, not fragmentation.

I’ve reached out to LangChain about collaboration. I have media relationships from my years as CTO of The New York Times and Chief Product & Technology Officer at The Wall Street Journal and Hearst that could help both frameworks reach enterprise audiences. They have developer community reach I don’t have. Together, we could move faster than separately.

The historical pattern

This has happened before.

DevOps emerged when multiple organizations independently recognized that development and operations needed to work together. Agile emerged when practitioners at different companies noticed the same problems with waterfall. The pattern: pain becomes obvious enough that people start building solutions, often simultaneously.

We’re in that moment for AI-assisted development. The gap between “vibe coding” demos and production-ready systems has become obvious. Multiple approaches are emerging to bridge it.

The fact that LangChain sees the same thing I see — that production AI development requires systematic practices, documentation patterns, and quality frameworks — validates that this discipline is real. It’s not one person’s opinion. It’s an observable need that multiple practitioners have independently identified.

Where to start

If you’re exploring these ideas:

• Synthesis Engineering covers the broader discipline
• Synthesis Coding focuses on the hands-on craft
• LangChain’s Agent Engineering addresses agent-specific systems

Read all of them. The perspectives complement each other. Take what’s useful. Contribute what you learn.

The industry is still figuring out what production AI development looks like. The more people engage with these frameworks — whatever we end up calling them — the faster we’ll converge on practices that work.

I’m looking forward to seeing where this goes.

Connect with me on LinkedIn if you’re thinking about these problems. I’m always interested in how teams are approaching production AI development.