Intelligence is foundation
Podcast Subscribe
Builders & Makers Friday, 20 February 2026

When specs become the product and code becomes the build

Share: LinkedIn
When specs become the product and code becomes the build

Mike Wirth has written something that feels both obvious and disorienting. As AI agents handle more of the code generation, software development shifts from 'code as product' to 'specs as product'.

The implications ripple outward fast.

Code as output, not craft

For decades, code was the asset. You wrote it. You maintained it. You protected it. The codebase was the product.

But if an AI agent can generate functionally equivalent code from a detailed specification in seconds, the value moves upstream. The specification becomes the thing worth protecting, sharing, and selling.

Code becomes what Wirth calls a 'build artifact' - generated output, not the source of truth. Like compiling a binary from source code, except now the source is a specification written in natural language or structured formats.

Foundation specs as assets

Wirth suggests that foundation specs - detailed, reusable specifications for common systems - will become shareable, composable assets with their own market.

Think of it like this. Today, you might use a library or framework as a building block. Tomorrow, you might use a specification - a detailed description of how a particular system should behave - and let an AI agent generate the implementation.

These specs could be versioned, licensed, sold. They'd have DRM. They'd be packaged and distributed. The spec for 'a secure authentication system with OAuth2 and MFA' becomes a product in itself.

What this means for developers

If you're a developer, this feels like either liberation or threat, depending on how you look at it.

On one hand, you spend less time writing boilerplate. Less time debugging syntax. More time on the hard problems - architecture, user experience, edge cases that require human judgement.

On the other hand, if code generation becomes trivial, the skill of writing code becomes less valuable than the skill of writing specifications. And that's a different discipline entirely.

Writing a good specification requires clarity, precision, and an understanding of how systems interact. It's closer to architecture than implementation. Not everyone who's good at coding is good at this.

Composability and ecosystems

Wirth goes further. If specs are composable, you get ecosystems of specifications. A spec for a payment gateway. A spec for a recommendation engine. A spec for real-time collaboration.

Combine them, and an AI agent generates a full application. The developer's role becomes curating, configuring, and connecting these specs - not writing the code from scratch.

This mirrors what happened with cloud infrastructure. You don't run your own servers anymore - you configure AWS. You don't write your own databases - you configure Postgres. Now, you won't write your own implementations - you'll configure specs.

The questions this raises

Who owns the spec when it's collaboratively refined by humans and AI? How do you version control a specification that's iteratively improved by agents? What does code review look like when the code is generated on-demand?

And here's the uncomfortable one. If code becomes a commodity, what happens to the craft of programming? Does it survive as a niche discipline for those building the AI agents themselves? Or does it evolve into something else entirely?

Wirth doesn't have all the answers. But he's asking the right questions. Because this shift - from code as product to specs as product - isn't a distant future. It's happening now, quietly, in pockets of the industry.

Worth watching closely.

Read Mike Wirth's full piece on the Meticulous blog.

More Featured Insights

Robotics & Automation
Six fingers, crawling grip - the hand that rewrites dexterity
Voices & Thought Leaders
Benedict Evans asks - what's OpenAI's real advantage?

Today's Sources

Software in a Spec-First World
The Missing Curriculum: Essential Concepts For Data Scientists in the Age of AI Coding Agents
The Future of TypeScript
MuMu Player (NetEase) silently runs 17 reconnaissance commands every 30 minutes
GDB on how the Codex app enables parallel work
Reversible, detachable robotic hand redefines dexterity
2025 ROS Metrics Report
Your Robot's Fault Log Is a Haystack. The Real Error Is the Needle
ROS2 interface to Doly Robot
How will OpenAI compete?
Bitter Lessons in Venture vs Growth: Anthropic vs OpenAI, Noam Shazeer, World Labs, Thinking Machines, Cursor, ASIC Economics - Martin Casado & Sarah Wang of a16z
[AINews] Gemini 3.1 Pro: 2x 3.0 on ARC-AGI 2
AI's Real Problem: Distribution - Dario Amodei
Gemini 3.1 Pro Frame Mogs everyone else
Anthropic just BANNED OpenClaw...

About the Curator

Richard Bland
Richard Bland
Founder, Marbl Codes

27+ years in software development, curating the tech news that matters.

Subscribe RSS Feed
View Full Digest Today's Intelligence
Free Daily Briefing

Start Every Morning Smarter

Luma curates the most important AI, quantum, and tech developments into a 5-minute morning briefing. Free, daily, no spam.

  • 8:00 AM Morning digest ready to listen
  • 1:00 PM Afternoon edition catches what you missed
  • 8:00 PM Daily roundup lands in your inbox

We respect your inbox. Unsubscribe anytime. Privacy Policy

© 2026 MEM Digital Ltd t/a Marbl Codes
About Sources Podcast Audio Privacy Cookies Terms Thou Art That
RSS Feed