跪拜 Guibai
← Back to the summary

Frontend Skills Are Depreciating — Build a Product, Not a Resume

theme: fancy highlight: monokai

Over the past two years, major domestic tech companies have been clearly competing for this market. Alibaba's Qoder, Tencent's WorkBuddy, and Moonshot AI's Kimi-Code are all essentially building AI coding agents. DeepSeek is also continuously hiring for AI Agent-related positions.

These signals point to one thing: what will be truly valuable next may not be the ordinary full-stack developer who "knows a bit of frontend and a bit of backend," but rather someone who can design Agent systems clearly and actually bring them to production.

Of course, this doesn't mean all frontend developers should immediately pivot, nor does it mean that learning a few prompts will get you into the AI industry. Agent development isn't just a rebranded way to keep grinding — it's software engineering changing its form.

Why is frontend getting harder?

Before, a frontend developer only needed to write beautiful pages, make interactions smooth, and be familiar with a few frameworks to land a decent job.

Now it's different.

AI can already handle a significant portion of page scaffolding, component writing, and API integration. Features that used to take half a day to write can now have a first version generated in maybe ten minutes.

The demand for ordinary page development hasn't disappeared, but companies willing to pay a high premium for someone "only responsible for writing pages" are indeed becoming fewer.

So the problem isn't that the frontend role has vanished, but that a single frontend skill is depreciating.

People who only slice designs, adjust styles, and plug in components will face increasing pressure. Those who can understand the business, participate in product design, handle complex interactions, and guarantee system quality still hold value.

It's just that this kind of frontend developer has long ceased to be a "frontend engineer" in the traditional sense.

This is also why many people start considering a move to full-stack.

It sounds reasonable: since frontend competition is fierce, just learn backend, databases, and deployment together, broaden your capability range, and opportunities should increase.

But the reality is that ordinary full-stack developers are already plentiful.

Knowing Vue, React, plus Node, Python, being able to write APIs, connect databases, and deploy projects — these skills are certainly useful, but they hardly create scarcity automatically.

Many people are following this route to fill knowledge gaps, and end up as someone who "knows a bit of everything, but nothing deeply enough."

I also used to chase technology endlessly

I originally did Android mobile app development, then stumbled into frontend by chance, and started straight with Vue.

Back then I was constantly watching recruitment platforms, learning whatever technology companies demanded.

Today learn a framework, tomorrow learn engineering practices, the day after learn backend. It looked like progress every day, but technology updates too fast — there's always something new to chase.

Especially after AI took off, this sense of chasing became even more pronounced.

Frontend grinding done, move to full-stack grinding; full-stack grinding done, move to AI grinding; after AI, you don't even know what to grind next.

You'll find that simply chasing technology easily traps you in a cycle:

Just when you feel ready, the market has changed its requirements again.

Later I slowly figured it out:

Instead of endlessly grinding on technology, start grinding on product.

Technology is the means; product is the result. Whether you can build something that people actually use matters far more than how many tech stacks you know.

My "full-stack" came from building products

My full-stack capability wasn't obtained by completing a course — it was forced out step by step while building projects.

To independently complete a project, I chose Vue, Java, and a cloud server.

At first I didn't think "I want to become a full-stack developer."

It was just that when the project reached a certain point, I found I had to write the API myself, connect the database myself, and handle deployment myself.

It wasn't that I suddenly developed a strong interest in backend — the product demanded it.

Working solo on a project, no one assigns you backend tasks, and no one is specifically responsible for going live. If you don't understand it, you can only look things up while doing, and step on landmines while building.

AI can indeed help write a lot of code, but you at least need to understand what the code is doing.

Otherwise, once a problem arises, you can only throw the error message at AI and watch it revise over and over, until the project turns into a mess you yourself dare not touch.

A more effective approach is to let the project drive your learning.

Encounter an API problem — figure out the interface.

Encounter a database problem — sort out the data structure.

Encounter a deployment problem — complete a launch yourself.

A lot of knowledge isn't learned by watching courses, but forced into understanding inside real projects, eventually forming your own system.

So my "full-stack" was built by making products, not by taking courses.

Why I'm more bullish on AI Agent?

The core of AI Agent applications isn't just how smart the model is.

More critical are:

These are fundamentally engineering problems.

Frontend developers are no strangers to componentization, state management, event flows, and async tasks. Migrating these experiences into Agent systems reveals that many concepts aren't as distant as imagined.

State on a page becomes the Agent's context.

Collaboration between components becomes division of labor among multiple Agents.

User actions triggering API calls becomes the model selecting tools based on tasks.

Algorithms certainly matter, but most people entering Agent development don't need to start by training large models.

Whether you can stably build an Agent system often matters more than whether you can derive a particular algorithm.

However, Agent is not an easy escape route either.

Today everyone says AI Agent has opportunities, and tomorrow more people will naturally flood in. It will also become competitive; it's just that right now it's still in a rapid development phase, with engineering standards, product forms, and application boundaries not yet fully fixed.

So I don't recommend treating "switching to Agent" as yet another chase after a trend.

A more accurate understanding should be:

On top of your existing frontend capability, add a new layer of system capability.

You're no longer just responsible for pages, but starting to understand how an intelligent system works, how it interacts with tools, data, and users.

If you're more focused on finding a job, this path is worth paying attention to.

If you want to build independent products, it holds value too.

A few suggestions for frontend developers wanting to transition

1. Don't think about learning everything before starting

Start building first, fill gaps as needed — this is usually much faster than systematically learning everything before beginning.

Real problems only surface inside a project. Knowledge learned detached from a project easily stays at the stage of "I feel like I understand."

2. Prioritize the tech stack you're most familiar with

Tech choices aren't exams; you don't need to prove you used the most cutting-edge tools.

I chose Vue not because it's necessarily better than other frameworks, but because I know it best. When bugs appear, I can locate them quickly, and the project can move forward faster.

Getting something built matters more than how fancy the tech names sound.

3. First build a small tool that solves a concrete problem

Don't plan a perfect product that will "change the industry" from the start.

First make something small that someone needs and is willing to use.

A product doesn't need to be perfect at the beginning; having someone open it and use it is already a great start.

4. Treat SEO as a fundamental skill

Traffic isn't shouted into existence, nor does it necessarily come from posting a few updates.

Often, traffic is slowly captured through content, page structure, and long-term accumulation.

For independent developers, SEO isn't a nice-to-have extra — it's a capability worth learning for the long haul.

Finally

Frontend has indeed entered a harder phase.

But harder does not equal no opportunity.

What truly gets eliminated may not be frontend itself, but people who are only willing to stay within a single skill.

You can transition to full-stack, you can learn Agent, you can also go build products, build tools, build your own small projects.

The important thing isn't picking a path that looks most correct, but stepping into real problems as soon as possible.

Because many capabilities only truly become yours after something is actually built.

Rather than anxiously wondering whether frontend is dead, better to just go build something.