AUTHORITY
Code Is a Commodity. Assets Are Leverage.
When one engineer and an AI can replicate a $500M framework in a week, the only defensible positions are the things AI cannot generate: owned properties, real readership, and editorial coverage.
One engineer. One AI. One week. $1,100.
In late February 2026, Cloudflare engineering director Steve Faulkner used Anthropic’s Claude and roughly $1,100 in tokens to rebuild 94% of the Next.js API surface from scratch. He did it in a single week.
The result, called vinext, is not a wrapper. It is a clean reimplementation of routing, server rendering, React Server Components, server actions, caching, and middleware — built on Vite rather than Vercel’s proprietary Turbopack. Early benchmarks showed builds running up to 4x faster and client bundles 57% smaller.
Next.js powers millions of production sites. Vercel has raised over $500M in venture capital. A single person with a model replicated the core in seven days.
The fallout reinforced the thesis
Vercel’s response was defensive. CEO Guillermo Rauch pushed back publicly. Vercel’s security team disclosed seven vulnerabilities in vinext within forty-eight hours.
AI can generate code that passes 1,700+ tests and 380 end-to-end validations. What it cannot generate is the years of hardening that come from real production traffic, real incidents, and real users hitting real edge cases.
Vercel’s counter-move — publishing a migration guide from Cloudflare to Vercel — was the tell. When your reply to a one-week clone is “we found bugs in it,” the codebase was never the moat.
What still has value
If a $500M-backed framework can be functionally replicated for $1,100, what is code worth? Code is now an execution detail. The things one engineer with a model cannot reproduce in a week are a short list:
- Owned properties. Premium domains are finite. AI creates more demand for addresses, not more addresses.
- Real readership. First-party audiences, organic discovery, years of behavioral data. These cannot be synthesized.
- Editorial coverage and citations. Trademarks, editorial trust, licensing rights, community relationships. Built over years.
The asset-first thesis
If you are building in 2026, the question is no longer “can we build the software?” The answer is almost always yes, faster and cheaper than ever. The question is now: what do you own that cannot be rebuilt in a week?
Premium domains are the clearest example of an asset that becomes more valuable as code becomes cheaper. Every AI agent, every auto-generated site, every new digital business needs an address. The supply of memorable, brandable .com domains is fixed. The demand is accelerating.
Where Pillar sits in this picture
Pillar Media & Entertainment operates the kind of inventory this thesis describes. We manage or are actively acquiring 100,000+ premium domain properties across English, Spanish, French, and Portuguese — the launch languages of our 6,608-domain inventory. The network reaches 500M+ monthly readers. Fifteen years of operating history sit behind it.
That is not a portfolio of code. It is a portfolio of the assets code is built on top of: finite addresses, real readership, editorial coverage, citations earned over time. Each property is a unit of leverage in exactly the era this article describes.
Code is abundant. Frameworks are replaceable. Platforms are contested. Assets compound.
The builders who win this decade will not be the ones who write the best code. They will be the ones who own what the code is built on top of.
Frequently asked questions.
What happened between Cloudflare and Next.js in February 2026?
Cloudflare engineer Steve Faulkner used Anthropic’s Claude to rebuild 94% of the Next.js API as vinext in one week, spending roughly $1,100 in tokens. It deploys to Cloudflare Workers, with benchmarks showing 4x faster builds and 57% smaller client bundles.
Does AI-generated code make software less valuable?
Yes. When a model can replicate an entire framework in a week for $1,100, code itself becomes a commodity. Durable value shifts to what code cannot replicate: owned domains, proprietary data, real readership, editorial coverage, citations, and licensing rights.
What are durable digital assets in the AI era?
Premium domain names, proprietary datasets, real organic readership, licensing rights to content and IP, and trusted editorial relationships. These assets appreciate as AI makes code and content production cheaper.
Why are domain names more valuable when AI commoditizes code?
Domain names are finite, non-replicable. AI can generate unlimited code, sites, and content, but it cannot mint new .com addresses. The fixed supply of memorable, brandable domains becomes more scarce and more valuable.
How is Pillar positioned for this shift?
Pillar manages or is actively acquiring 100,000+ premium domain properties reaching 500M+ monthly readers across English, Spanish, French, and Portuguese. That inventory is the exact asset class this thesis describes.