Pipe Dream.
Matt Shumer's piece has been making the rounds, arguing that AI has made "building" absurdly fast (He's not wrong and and it's a worthy read).
I buy the premise. I've seen the demos. Heck, I have a GitHub account with a growing collection of repos I built with these tools. A thing that used to take a sprint now shows up before your second coffee. I'm on both sides of this: genuinely thrilled by what I can create, genuinely sober about what happens next.
And look, I'm no Luddite, but shipping in the enterprise is where the genre shifts.
Identity. Access. Provisioning. CI/CD. Security review. Observability. Cost controls. Governance committees whose calendars move at the speed of Outlook invites (well behind inference tokens). The parts of the company designed to keep production safe did not speed up just because the code did, and the parts of the company designed to keep the business compliant aren't even sure what question to ask yet.
Think about it like oil. We talk about oil as though it's a fuel problem, cars and gas stations, easily swapped for electrons. But oil is also plastics, synthetic textiles, pharmaceuticals, fertilizer, asphalt. It is woven so deeply into the substrate of modern life that "just stop using it" lands somewhere between bumper sticker and fantasy. We may never rip sweet crude pipelines out of the ground, but ripping out tried-and-true systems of governance, commerce, and psychological infrastructure? That's going to take time. Good thing nobody can get their hands on memory I suppose.
That is roughly where enterprises sit with their legacy delivery systems. AI is the new energy source: absurdly powerful, increasingly cheap, available to nearly everyone. But organizations don't run on energy. They run on pipes. And those pipes were cast, cooled, and certified for a completely different fuel.
So you get a two-speed organization. Prototypes sprint. Production crawls. Everyone talks about acceleration while release calendars quietly remain … Tuesday.
The move now is treating the delivery system like a product: paved roads, self-serve environments, policy-as-code guardrails, repeatable releases, and incremental modernization that lets new capabilities earn their way into the core.
Because the moat is shifting toward whoever can ship safely, repeatedly, and without turning every release into a round of multilateral diplomacy.
Recommended by LinkedIn
And here's the part nobody's pricing in yet: the builders aren't just engineers anymore. There are now MILLIONS of people with GitHub repos, most of whom have never taken a computer science class. MILLIONS of prototypes looking for a production lane. If you want a preview of where this lands, imagine two fourteen-year-olds a few years (months?) from now:
🪩 Zuri: yo fractal-flux beat app ↑ on GH. mood-coded + deployed.
🧠 Kai: send link. free grab?
🪩 Zuri: dl free; coin-toss a few bucks on Venmo tho … rent-boss energy 😵💫
🧠 Kai: lol bet. pulling now. 🚀
🪩 Zuri: sync waveforms + drop PRs. peace. ✨
That's a Tuesday.
Anyway, off I pop to show my Mom how to deploy the iPhone app she vibe-coded to the App Store. It's called "What's This?" You point the camera at stuff and it says things like, "This is a dog, not a remote for the TV.
The future is already here. It's just unevenly deployed.