Stop Reading About AI. Start Shipping With It.

TL;DR

The volume of what a single builder can ship with a tuned AI coding environment has fundamentally changed the economics of starting something
AI skills will inevitably commoditise, and the narrow window where these capabilities give you outsized leverage is open now and closing
Stop consuming content about AI and start producing things with it. The learning loop only works when you're building

I need to say something directly to the people in my network who are still in observation mode.

Stop reading. Start shipping.

I don't mean stop learning. I mean stop treating learning as a substitute for doing. Stop reading the fourteenth article about how AI is changing everything. Stop watching the demo reel. Stop bookmarking the tutorial you'll get to next weekend. Stop waiting for the perfect moment, the perfect tool, the perfect idea.

The world has fundamentally changed. I've written about it from every angle: the builder-leader identity, the PM-to-builder shift, the collapse of build economics, the team of one. But writing about it and living it are different things. And I can now say with absolute confidence, from the other side of shipping production AI systems as a solo operator, that the sheer volume of what you can accomplish with a tuned IDE connected to a capable AI coding agent is unfathomable.

Not theoretically. Not in a demo. In practice. Shipped. Running. Serving users.

Your title is irrelevant

The mindset shift that unlocks everything: your past experience does not dictate your future output.

I spent years as a Director of Product at an enterprise company. That title, that experience, that trajectory: none of it determined what I could build when I sat down with AI tools and started shipping. The constraint wasn't my background. It was the decision to start.

The same is true for you. If you're a product manager who's never written production code, you can now build and ship a functional product. If you're a designer who's always depended on engineering to bring your vision to life, you can now prototype and validate independently. If you're a domain expert in real estate, finance, healthcare, or legal who's always had ideas but never had the technical capability, the barrier is gone.

Not lowered. Gone.

I'm not being hyperbolic. The combination of AI coding agents, modern deployment platforms, and pre-built infrastructure means that a single person with domain expertise and product taste can build, deploy, and scale a vertical SaaS product that would have required a team of ten three years ago.

If you see a gap in the market for a high-end service platform in your domain (a better tool for property managers, a smarter workflow for paralegals, a more integrated system for procurement teams) just start building. The technology will meet you where you are.

The window is narrow

This is the part people don't want to hear.

Right now, AI-augmented building skills are a genuine competitive advantage. The people who can use AI coding agents effectively (who can shape problems clearly, curate context, evaluate output, and iterate rapidly) are operating at a productivity level that the market hasn't priced in yet.

But this window will close.

AI skills will inevitably commoditise. Today's power-user capability will become tomorrow's baseline expectation. The tools will get easier. The workflows will get standardised. The training programmes will proliferate. Within two to three years, the ability to build with AI will be as expected as the ability to use a spreadsheet.

When that happens, the advantage shifts from "can you use AI to build?" to "did you build something that compounds?" The person who ships a vertical SaaS product this year, who captures users, accumulates data, builds integrations, and establishes trust, has a compounding advantage over the person who starts the same project two years from now, even if the later entrant is equally skilled.

The next generation of vertical SaaS won't be built by armies of engineers at well-funded startups. It will be built by individuals with domain expertise who moved first. The three skills that define the modern PM (problem shaping, context curation, and taste) are available to anyone willing to start. Individuals who saw the gap, picked up the tools, and shipped while everyone else was still reading about the future.

First-mover advantage has never been more accessible or more perishable. Act before the market saturates.

The recursive learning loop

The part that surprises people who haven't started yet: the fastest way to learn AI is to build with AI.

Not to study it. Not to take a course. Not to read a whitepaper. To build something real, with real users, encountering real problems.

Use AI to learn AI. When you don't understand how a particular API works, ask an AI coding agent to explain it while building the integration. When you're unsure how to structure your data model, have the agent generate three approaches and evaluate them. When you hit a bug you don't understand, debug it with the agent as your pair programmer.

Use AI to build AI. When your product needs an AI feature (a classification system, a recommendation engine, a natural language interface) build it using AI tools. The recursive loop of using AI to build AI products is the fastest skill development path that exists. You learn the capabilities and limitations simultaneously, because you're encountering them in the context of a real problem with real constraints.

Use AI to master AI. Every shipping cycle teaches you something that reading can't. How to write better context documents. How to evaluate agent output faster. How to shape problems more precisely. How to recognise when the agent is confidently wrong. These skills only develop through repetition, and the only way to get repetition is to build.

The learning loop is: build, ship, learn, build better. Not: read, plan, prepare, eventually build. The preparation phase is a trap. The building phase is where the learning happens.

What this actually looks like

I want to be concrete about what's possible, because abstract inspiration doesn't ship products.

A single builder with domain expertise and AI coding tools can, today:

Build a full-stack web application with authentication, database, API layer, and polished frontend in days, not months
Deploy it to production with CI/CD, monitoring, and custom domain in hours
Iterate based on user feedback with same-day turnaround on feature requests
Add AI-powered features (search, classification, recommendations, natural language interfaces) without a machine learning team
Scale to hundreds of users on modern infrastructure without ops expertise

Three years ago, each of those bullets required a specialist. Today, they require a builder with a clear problem and the willingness to start.

I'm not suggesting that solo-built products are automatically as robust as what a well-resourced team produces. They're not. Production-grade security, enterprise compliance, and sophisticated data engineering still benefit from deep expertise. But the gap between "solo builder prototype" and "viable commercial product" has compressed dramatically. For many vertical SaaS use cases, a solo builder can ship something that's genuinely competitive.

The call

I've spent months writing about the theoretical and strategic implications of AI for product builders. The frameworks matter. The architectural thinking matters. The economic analysis matters.

But none of it matters if you don't build.

The hype is real. The fear is escalating: fear of being left behind, fear of AI replacing roles, fear of the pace of change. The opportunity is genuine and time-limited.

So: what are you shipping today?

Not what are you reading. Not what are you planning. Not what are you considering. What are you shipping?

If the answer is "nothing yet," that's fine. But make it the last day that's true. Open your IDE. Connect an AI coding agent. Pick the problem you know best. Start building.

The world changed. The tools are ready. The only thing missing is your decision to start.

Frequently Asked Questions

What if I don't have a product idea?

You do. You just haven't framed it as one. Think about the most tedious part of your current job: the workflow that frustrates you, the manual process that wastes your time, the information gap that slows your decisions. That frustration is your first product idea. Build the tool you wish existed. If it solves your problem, there's a high probability that others in your domain have the same problem.

How do I know if my AI-built product is good enough to ship?

Ship it before you think it's ready. The standard for "good enough" is: does it solve one problem for one user better than the alternative (which is often a spreadsheet or manual process)? If yes, ship it. Get real feedback from real users. Iterate. The AI-built product you ship today and improve tomorrow will always beat the perfect product you're still planning next month.

Won't the market be saturated with AI-built vertical SaaS?

Eventually, in the broadest categories. But vertical SaaS is called vertical for a reason. The niches are nearly infinite. The closer your domain expertise is to the problem, the harder it is for a generalist builder to compete. A property manager who builds a property management tool has insights that no outside builder can replicate. Domain expertise plus building capability is the winning combination, and the number of people who have both in any given niche is very small.