Yes, programming is still worth learning if you want to build software you can understand, change and maintain. AI can help you get started, explain unfamiliar code and suggest solutions. The useful skill is being able to decide what should happen, check what actually happens and fix the difference.
If your only goal is a simple website for your own business, you may not need to become a developer. If you want to develop applications professionally, learning how the code works remains a worthwhile investment. These are different goals, and your learning plan should reflect yours.
I am a frontend developer, and I love AI. Here is the learning route I would recommend to someone starting frontend development now.
Use AI to help you learn, then check what you learned
There is a useful warning in Anthropic’s January 2026 study of coding skill formation. In a small randomized study of developers learning an unfamiliar Python library, the AI-assisted group scored lower on a subsequent knowledge assessment. Participants’ approaches to using AI differed, and some used it to build understanding. This was a specific learning exercise, not evidence that every learner or AI tool produces the same outcome.
My practical response is to separate two activities: getting the feature working and checking whether I can explain it. A completed feature is evidence of output. Explaining its behavior, changing a requirement and fixing a new bug are better checks of understanding.
Use AI freely for questions. Occasionally close the chat and repeat a small part of the task yourself.
Choose a direction before choosing a stack
For frontend development, start with HTML, CSS and JavaScript. MDN’s curriculum provides a structured foundation that also includes accessibility and version control, with frameworks and other specialist topics in its extensions.
You do not need to finish every tutorial before making something. Pick a small project and learn the next concept when the project needs it.
| What to practise | A concrete task | How to check your understanding |
|---|---|---|
| HTML and forms | Add a company, role and status to an application form | Explain each input’s label and purpose |
| CSS and layout | Make the form usable on a narrow screen | Fix an overflow problem without replacing the stylesheet |
| JavaScript and state | Add, edit and filter applications | Explain which data changes after a click |
| Browser debugging | Investigate an incorrect application count | Find the first incorrect value |
| Git | Save each working feature as a commit | Describe and undo one change |
| A framework | Rebuild one view after understanding the original | Explain what the framework simplifies |
Choose Vue or React when you have a reason to use it. Avoid learning both simultaneously in your first project. Add TypeScript after you can explain the JavaScript you are writing; give important data structures explicit types instead of treating type errors as something to silence.
Build one project through six increasingly useful versions
A job-application tracker is a suitable practice project: it has forms, lists, changing state and decisions about user experience. Use fictional applications while building it. The following sequence is my suggested exercise, not a course with a promised completion date.
1. Make a usable static screen
Create a heading, an application form and a short list of example entries. Include a company, job title, status and application date. Keep the design simple enough that layout does not consume the entire project.
Before moving on, navigate the form with a keyboard and check it on a narrow screen. Write down one problem you found and how you fixed it.
2. Add behavior with ordinary JavaScript
Allow someone to add an application and change its status. Decide what happens when a required field is empty. Start with data held in memory so you can concentrate on the relationship between the data and the screen.
Explain the journey aloud: the form is submitted, values are read, an entry is created and the visible list changes. If one step is unclear, investigate it before generating more features.
3. Preserve and filter the data
Add local saving as a learning exercise, then filter by status. Define what should happen when no entries match. Decide whether a counter shows all applications or only the filtered results, and label it accordingly.
Now introduce a requirement change: one application can have more than one interview. Your first data model may need to change. That is a useful part of the exercise, not a failed first attempt.
4. Practise loading and failure states
Replace the in-memory list with a supplied local JSON file or a small mock API. Build separate states for loading, an empty result and a failed request. Give the user a useful next action after a failure.
Deliberately break the request URL. Inspect what happened in the browser, repair it and explain why your error message appeared. Keep real credentials and personal applications out of this demo.
5. Refactor with a clear reason
Introduce your chosen framework and split the interface into a form, a list and a status filter. Compare the new version with the earlier one. Which responsibility became easier to locate? Which abstraction made the code harder to follow?
Write a few checks for meaningful behavior: an empty company name is rejected, changing a status updates the list, and the empty state appears for a filter with no matches. Avoid counting tests as a substitute for choosing useful cases.
6. Make the project explainable
Publish a demo with fictional data and write a short README. Describe the problem, the main decisions, how to run it and what remains unfinished. Explain one bug in enough detail that another developer could reproduce it.
An explainable project gives you something concrete to discuss. You can show a requirement, your implementation and the reasoning behind a tradeoff.
A 45-minute practice session
Use this as an adjustable routine:
- Five minutes: choose one behavior, such as editing an application without duplicating it.
- Fifteen minutes: attempt it yourself and note where you get stuck.
- Ten minutes: ask AI one focused question, then inspect its suggestion.
- Ten minutes: try normal and awkward inputs, and check the browser output.
- Five minutes: explain the change and record the remaining uncertainty.
The times are a planning aid, not a learning rule. If the same concept keeps blocking you, spend the next session on a smaller exercise. For example, practise updating objects in an array before returning to an entire application editor.
Ask for help that leaves you thinking
Try this prompt when you are stuck:
I am learning JavaScript. My goal is: [one behavior].
Expected result: [what should happen].
Actual result: [what happens instead].
Here is the smallest relevant code: [code].
First ask me to predict what this code does. Then identify one
concept I may be misunderstanding. Give one hint at a time.
Do not rewrite the application or add dependencies.
After I fix it, give me a similar small task to solve myself.
For example, suppose the status filter works but the application counter looks wrong. Check which array the counter reads. Is it counting every application while you expect only visible ones? That is a specific question about data and requirements. Asking AI to “fix the dashboard” makes the task much harder to inspect.
You can also ask AI to propose explanations for a bug. Choose one explanation, identify the evidence that would support it and check that evidence in the browser. Treat the response as a hypothesis until the behavior confirms it.
Decide whether you are making progress
At the end of a project version, ask yourself:
- Can I explain the path from user action to updated data and screen?
- Can I change one requirement without generating the application again?
- Can I investigate an error using the console, network panel or a breakpoint?
- Can I describe a limitation and what I would change next?
If you cannot yet answer something, you have found your next lesson. You do not need to hide that gap behind a larger project.
For a job search, present this evidence alongside your experience and learning. A project cannot promise employment, but it can make your current abilities much easier to assess.
Use the learning-project checklist to record what works, what you can explain and what you want to improve next.
Frequently asked questions
Should I learn programming if AI can already write code?
Yes, if you want to make and maintain software. Learn enough to specify behavior, read the implementation and investigate failures. If you only need a one-off personal page, using a website builder may be a better fit than beginning a developer career path.
Which programming language should I learn first?
For frontend web development, start with JavaScript alongside HTML and CSS. For another goal, choose the language used by a small project you actually want to build. Keep the first project narrow enough that language switching does not become your main activity.
Should beginners use AI while learning?
They can. Ask for explanations, examples and hints, then check understanding with an independent attempt. If you repeatedly accept code you cannot explain, reduce the task and ask for help with the underlying concept.
How long will it take to become job-ready?
There is no reliable fixed number of weeks. Your starting point, available time, target role and local market differ. Assess progress through work you can explain and modify, and compare that evidence with actual vacancies rather than a course completion badge.
Related reading: How to Tailor Your CV with AI Without Inventing Experience · Claude Code vs Codex for Frontend Development: A Practical Comparison · Which Jobs Will AI Replace? Jobs at Risk in 2026
Sources
- MDN Curriculum — Mozilla
- How AI assistance impacts the formation of coding skills — Anthropic, 29 January 2026
Building a feature for your web application?
The useful part of a technical approach is how it works for your users. I build interfaces, connect them to existing APIs and help product teams improve features, including loading, empty and error states.
Explore frontend development for your application
Related work and examples
Dashboard
A COVID data explorer grew into a personal dashboard and a collection of everyday web tools. A long-running React project where I learned to connect interfaces, interactive charts and real data.