A Python textbook went unread and was sold. The website still got built. A firsthand look at delegating coding, testing and releases to AI, with a human developer as backup.
A few years ago I bought a Python textbook because programming sounded interesting. I barely opened it. Eventually I sold it.
Today I run a website. I have not personally written its code, executed its tests by hand, or carried out the technical release steps. I decide what should exist, let AI work on it, and evaluate the result.
Can someone build and operate a website without learning to code? For some new projects, yes. AI and automated systems can do much of the technical work. The remaining job is to define tasks, check results, and know when specialist help is worth paying for.
The textbook was sold. The website kept growing.
The conventional route is to study programming, write code, and then build a product. My route began with an idea. I described a feature to AI, asked it to implement the feature, checked the result, and requested changes. That cycle eventually supported multilingual content and regular updates.
I still handle the decisions and account registrations that require the owner. I may not know the syntax of Python, but I do know what outcome I want.
Somehow I skipped the textbook chapter on variables and landed in the chapter called “Run the service.” The curriculum seems broken. The software, however, is real.
Even reading error logs became someone else's job
A good-looking page is not the same as a stable service. Changes can break other features. Tests fail. Releases stall. Then the screen fills with log messages.
My first thought was not “How fascinating.” It was “Do I really have to read all of this?”
I began asking AI to read the errors, investigate causes, make fixes and check again. Failures remain possible, but I no longer need to decipher every line myself.
Sometimes the limiting factor is the AI service's weekly usage allowance rather than my ability to write code. That describes one person's setup; it is not a guarantee that every old or complex system is easy to automate.
No ready-made kitchen? Hire the prep work.
In a discussion about software automation, somebody compared a well-prepared development environment to a kitchen where your mother has already bought and chopped the ingredients.
There is a real point here. Someone had to install the tools, connect services, and untangle old code. Teams maintaining inherited systems do not start with the same advantages as a brand-new project.
But someone who never wanted to be the chef asks a different question:
“If I need someone to prepare the kitchen, why don't I hire someone to do that?”
It sounds ridiculous. It is also a workable way to divide labor. Just as you can pay for help packing during a move, you can commission the technical setup from AI or a developer.
The difficulty does not disappear: you still need to find someone, agree on a price, and inspect the outcome.
When the AI hits a limit, there is a human backup
If AI runs out of usage allowance or repeatedly fails to solve a problem, you can wait or commission a developer to handle a clearly defined task.
Knowing someone who works in web development makes that option more tangible. Their relevant skills, availability, fee and willingness still need to be checked.
This is about paying fairly for the right expertise, not finding the cheapest possible person. Delegate the work the AI can do; use a specialist for specialist problems; retain decisions only the owner can make.
Three things the owner still has to check
First, open the actual public page. “Release completed” does not prove visitors can see the correct version.
Second, ensure a faulty change can be rolled back.
Third, increase scrutiny where payments, personal data or security are involved. Independent checks or expert review matter more when the downside is substantial.
GitHub's documentation explains automatic builds and tests after changes. Google's 2025 DORA research found that AI can amplify both existing strengths and weaknesses in software development. Automation can deliver mistakes quickly, too.
The cookbook is gone. The kitchen is operating.
Learning to code remains valuable, especially for people who enjoy it. But if the goal is a working website, personally writing every line is only one route.
I never finished the Python book. I don't want to read pages of logs either.
I didn't become the chef. I became the person who hires the kitchen—and may outsource the grocery shopping as well.
References (2)
- GitHub Docs, Continuous integration (accessed 2026-10-14): https://docs.github.com/en/actions/get-started/continuous-integration
- Google DORA, State of AI-assisted Software Development 2025: https://dora.dev/research/2025/dora-report/
