The New Code: Why Spec-Writing Is the Real Superpower in the Age of AI

In traditional software development, we’ve always known that clear specifications and requirements are essential for building valuable products. Now, as we work more with AI assistants and agents, this principle hasn’t changed—in fact, it’s become even more critical. Sean Grove, in his talk “The New Code,” makes it clear: writing code is only a small part of creating software, and arguably not the most important part.

One of the most interesting points Sean raises is that what we need to save, version, and protect is no longer just the source code, but the specifications themselves. Well-crafted specs become the real asset—if they’re written clearly, they allow us to generate the product in any language or technology we choose.

Sean, who works on alignment reasoning at OpenAI, lays out a compelling case for why specifications—not code or prompts—are becoming the fundamental unit of programming. Here are my key takeaways and favorite quotes from his presentation:

1. Communication trumps code

“Code is sort of 10 to 20% of the value that you bring. The other 80 to 90% is in structured communication.”

The real bottleneck in software isn’t just writing code, but understanding user needs, planning, sharing ideas, and verifying if what you build actually solves the problem.

2. Specs are the new source of truth

“Code itself is actually a lossy projection from the specification”

A good spec contains all the necessary requirements to generate code for multiple architectures and serves as the single source of truth that can be compiled into code, documentation, tests, and even podcasts.

3. Specs align humans—and machines

A written specification “effectively aligns humans” and becomes the artifact to discuss, debate, and synchronize on. It’s not just about aligning teams; executable specs can also train and evaluate AI models. Sean describes OpenAI’s Model Spec as a living document, “a collection of markdown files” that everyone—from engineers to legal and policy—can contribute to.

4. The future is about intent, not syntax

“Engineering is the precise exploration by humans of software solutions to human problems”

We’re moving away from machine-focused encodings toward a unified human encoding of intent. In this new era, whoever masters spec-writing will be the most valuable programmer.

So next time you start a project, don’t just jump straight into code. As always, begin with a clear, executable specification.

“Whoever writes the spec… is now the programmer.”

How is your team approaching specs in the age of AI? Let’s discuss!

Sources:

• youtube talk

• model specs