The AI problem for freelancers
The work clients used to hire you for is now available to everyone with AI.
That does not mean your experience is worthless. It means a decent first draft, landing page, content plan, workflow, or small site is cheaper and faster to get than it was. More people can do the basic work clients used to pay freelancers to do.
So the answer is not automatically more client hunting or lower prices. The better mission is to use the skills you already have to build useful projects for yourself. Hi, I’m Virgil. I am helping freelancers shift from working for clients to working for themselves.
The problem
When a client can get a usable first pass in five minutes, selling another first pass is a bad long-term plan.
AI is not replacing every freelancer. It is making routine output less scarce. If your income depends on repeatedly producing work that is now easy to generate, the pressure does not go away because you get better at proposals or agree to work for less.
You can still keep clients you like. But client work alone leaves you exposed to somebody else’s budget, priorities, and decision to use a cheaper tool or person next time.
The solution
Turn your freelancer skill set into projects you own.
Self-freelancing is not pretending you are a venture-backed founder. It is using the research, scoping, building, and shipping skills you already sell to create a small tool, guide, site, or service that you control when the work is done.
That gives you more than another invoice. You keep the asset, the audience relationship, the learning, and the ability to improve it without asking a client for permission.
Who this is for
You can deliver for clients. Now you need a project that keeps delivering for you.
This is for experienced freelancers whose old work is getting more crowded, cheaper, or harder to differentiate. You do not need to quit client work, call yourself a founder, or build a startup. You need one narrow project with a real person, a clear problem, and a finish line.
The longer you wait, the more of your best hours go toward assets that belong to other people. The right first project does not need to replace your income. It needs to prove that you can direct your skills at your own problem and finish something.
The method
Do not start by guessing. Start with a problem someone already took the time to explain.
I find a useful YouTube video, run it through YouTube to Transcript, and give the transcript to Codex. Then I ask: who has this problem, what is the smallest useful answer, and what does version one leave out?
The transcript gives you the language, friction, and questions. Codex helps you turn it into a scoped project. An Astro theme gives you a clean shell without making design busywork your new excuse.
The real objections
“I do not have time, I cannot code, and I might build the wrong thing.”
You are right to be careful. That is why the first project should be narrow: no big platform, no five integrations, and no six-month plan. Use a fixed block of time, a transcript from a real problem, Codex for implementation help, and a theme for the structure.
If it does not land, you did not waste a year. You learned what a real user did not need and built the muscle of finishing your own work. If it does land, you own the next decision instead of waiting for a client to make it.
Why listen
I have worked with clients around the world since 2004, and full time since 2023.
That is proof that I know how to take a loose brief, ship the work, and deal with the messy parts. It is not a promise that every owned project will work. I am making this shift myself, showing the work as it happens, and sharing the parts that do not work too. You can read the past client work that taught me that delivery discipline.
My difference is not generic AI advice. It is a repeatable project brief built from real-world language: video, transcript, scope, Codex, Astro, publish.