Last Sunday, between breakfast and midnight, I shipped: a complete marketing site for a new consultancy (opens in new tab), live on its own domain, with a working enquiry pipeline, an interactive product demo, structured data, and Google already indexing it. Plus a full repositioning of this site: a new advisory offer, a blog platform, an SEO and accessibility overhaul, and a conversion pass across every page.
One day. One person. No agency, no developer, no designer on retainer.
What that used to cost
I've commissioned this kind of work the traditional way, so the comparison isn't hypothetical.
A branded marketing site with custom design, eight pages, an interactive tool and a lead pipeline: a mid-tier Australian agency quotes $25,000 to $50,000 and delivers in eight to twelve weeks. A capable freelancer might do it for $8,000 to $15,000 over a month or two, with less polish on the edges.
The personal site overhaul, with the SEO layer, the accessibility pass and the new positioning, is another $10,000 to $20,000 and several more weeks, mostly spent in meetings about the meetings.
Call it $40,000 to $70,000 and three to four months of elapsed time, across two projects that would each have their own kickoffs, their own revision rounds, and their own invoices.
The Sunday version cost a domain name, some hosting that rounds to zero, and a day of my attention.
The stack, since you'll ask
Nothing exotic: Next.js for the sites, GitHub for version control, Vercel for hosting and deployment, Claude doing the heavy lifting on build and copy inside an agent-first IDE, a form relay for enquiries, and Google Search Console watching it all get indexed. I've written before about the tools I run across multiple companies. Every tool on that list is publicly available and most are free at this scale.
Which is the point. The stack isn't the advantage, anyone can install it by Tuesday. The advantage is knowing what to ask it for, in what order, and what to reject when it comes back. The tools have been democratised. The taste hasn't.
The part everyone gets wrong
The lazy conclusion is "AI makes websites cheap now." That's true and it's also the least interesting thing happening.
Websites were already cheap if you wanted cheap. What changed is the cost of iteration at quality. On Sunday, the site was reviewed and rebuilt several times before dinner: claims audited, copy tightened, pages restructured, a legal risk in the wording caught and removed sitewide. In the traditional model, each of those rounds is a change request, a week of turnaround, and a line item. The revision loop is where agency projects go to die, and it's the exact thing that has collapsed to minutes.
The other thing that changed: the floor didn't just drop, the ceiling didn't either. The output wasn't a template with my logo on it. Custom brand system, custom interactive tooling, proper structured data, accessibility done to standard. The trade-off between fast, cheap and good hasn't disappeared, but the triangle has been redrawn so aggressively that most people haven't updated their instincts yet.
What actually gets scarce
When execution stops being the bottleneck, the bottleneck moves. It moves to judgement.
Every hour I saved on building was spent on decisions the tools can't make: what claims a business in a regulated space can safely put on a page. What a fair description of shared work looks like when you have co-founders. Where a price should never appear. Which of three positioning angles earns the door and which one burns it. The building was fast because the deciding was deliberate.
That's the real Sunday test, and it has nothing to do with software. Give an operator one day and these tools, and the output is a business asset. Give the same day and tools to someone without the judgement layer, and the output is a very fast first draft of the wrong thing.
Why this matters if you run an SME
Every quote you're sitting on right now, for a website, a system, an internal tool, a process that "needs a platform," was priced against the old economics. Some of that work still deserves the old price. A lot of it no longer does, and knowing which is which is now a core commercial skill, the same way reading a P&L is.
The businesses that internalise this early get a compounding advantage: they test more ideas, kill bad ones cheaper, and ship the good ones months before their competitors' agencies have finished discovery.
The ones that don't will keep paying four months and five figures for a Sunday.
If the judgement layer is what your business is missing, that's what I do.