back to the blog

The last time my full time job title said web developer, it was 2016. While I kept building websites for freelance projects, continuing to maintain this site, I grew farther away from learning new techniques and understanding what obstacles face developers. 10 years later, as a senior technical writer for Google Chrome, my job is to write and advocate for web developers. I feel confident in my ability to do so, despite this distance from the role, but I want to accurately reflect our current technical landscape. With generative AI everywhere, it’s unsurprising to see job postings and expectations for what makes a web developer start to change.

As I work on the documentation strategy for Chrome’s AI platform, I’ve found myself asking the following questions:

• What skills does a web developer need in 2026? Will that be true in 2027 and beyond? Can we even really project at this rate of change?
• What roles are web developers be expected to take on? Do they need to be AI system architects? Do they need to be generative AI experts? Are they expected to have a dual role as ML engineer? Or, are they expected to leave the system questions alone, while still expecting to define what the web experience is for users.

To try and answer these questions, I took a look at the job market.

What are the expectations on the job market?

I took a look at current job postings for web developers and front-end engineers, to try and understand what companies are sharing about their AI-related expectations. Some posts are devoid of any mention of AI, while others expect an AI-augmented workflow. Some expect front-end engineers to build AI features and work with LLMs. Of course, it’s likely this will continue to change as companies change their internal tooling and priorities.

The pressure is on to use AI at work to make workers “more productive,” just read the news. Companies are calling for it publicly and privately, but has web development fundamentally changed?

AI tools versus AI development

With these expectations comes a common question: are developers expected to use AI as a tool to do work or expected to build AI features for others to use? The former may require significantly less machine learning knowledge, although understanding of how it works can make a tremendous difference your effectiveness. In Chrome, we’re betting that both are true. It’s why we published Learn AI for web developers.

Mike Masnick wrote AI Might Be Our Best Shot At Taking Back The Open Web. As a writer, what sticks out most to me is:

For all those years that tech bros would shout “learn to code” at journalists, the reality now is that being able to write well and accurately describe things is a superpower that is even better than code.

f you can tell a coding agent what to do in natural language, if the model has enough context of the language and common building techniques, you may be able to build faster, less error-prone applications. Does that mean technical writers are even better equipped to be developers? Or perhaps our partnership is more important than ever.

What happens to the web?

The key piece of the “AI can code” logic is that the model knows and understands the technology to be able to build something. This means it’s much more difficult to build with new APIs and frameworks. How are we supposed to keep improving the web if we build on older technology? Why would someone try to adopt a new technology if they have a their trusty coding companion that’s better at building with the old code, and their boss is breathing down their neck to use AI to go faster?

The companies building those new technologies (my employer included) rely on the hope that models will pick up on new technology faster and faster. If that comes to pass, maybe the web won’t be left to out to dry with sites built on aging technology.

There are more questions than there are answers. The major large language models have moved faster than many of us have imagined. A year ago, these questions felt preposterous.

Are these questions worth asking?

There are many soon-to-come costs as model providers move past the patience of investors who are waiting on their return. I remember the early days of Uber, which fundamentally changed how young New Yorkers got around town. Of course that pricing model couldn’t last. Free tokens aren’t forever. Will the idea of “AI is the solution to all” last when the bill comes due?

As demand for managed model providers surges, so does demand for more data centers. Is our water worth saving? The name for the cloud has broken people away from the fact there are very real places and environmental resources being maxed out, for the sake of compute power.

And, is it worth writing about this topic when our jobs are on the line? I’ve asked myself since joining Chrome’s AI efforts. I’m aware that even sharing these questions could have consequences. Will it be worth it?

I’ll keep asking questions. As we head into summer, I’ll have the opportunity to meet with developers from around the world, thanks to Google I/O and Google I/O Connect. I hope to hear about their experiences, gain insights that help my team better support developers efforts, and intuit a little more about what the future holds for the web.