The next two years of software dev...

I’m going to be honest, I literally have no idea how the future will look like for us devs. I don’t know if Claude 6.9 will actually replace me, just like you guys keep saying in the comments, and I don’t know if AI will end up being just another tool in our arsenal which helps us move faster at a fraction of the cost.

What’s more important is that I really do believe no one knows how this AI business will actually settle once the hype cycle burns off and the incentives become clear.

But, luckily for us there still are people in the industry who are trying to look at this from a realistic, objective perspective instead of echoing whatever AI tools went viral last week.

Adi Osmani spent years working on web performance and developer tooling at Google. He’s known less for hype and more for translating emerging trends into practical implications for developers, especially around productivity, performance, and the long-term direction of the web ecosystems. When he shares his take on the future of software development, we should all sit down and listen.

So in this video we’ll review how the next 2 years will look like for software developers through the eyes of a leading software engineer who currently works on Google Cloud and Gemini.

The truth is that the software industry has reached one of those moments that only becomes obvious in hindsight. For years,

the story was really simple. Software was eating the world, capital was cheap, and companies hired anyone who knew the difference between Java and JavaScript. Growth mattered more than efficiency and the solution for any problem was to simply hire more engineers.

But, by the looks of it, AI is breaking that assumption because productivity is no longer strictly linear with hiring, and that forces companies to rethink what an engineering team is supposed to look like.

Adi’s argument is not that AI replaces developers, but that it compresses the distance between idea and execution. The friction that used to justify large teams is shrinking because a lot of the boilerplate, setup and repetitive work needed in the past can now be automated quite successfully by some of these LLMs. These days, there are 5 unknowns when it comes to the impact of AI in the software development field.

[1. Are juniors in danger?] First of all, we all know that juniors and university students are going through a really rough patch because the traditional entry path into software development is no longer stable.

For decades, the career path was straightforward. Learn the fundamentals, get hired into a junior role, gain experience, and gradually move up the ladder.

AI weakens the economic logic that supported this pipeline, and early data already points in that direction.

Studies tracking labor adoption patterns show that when companies introduce generative AI into engineering workflows, junior hiring declines while senior hiring remains relatively stable.

When a senior engineer assisted by AI tools can complete tasks that previously required multiple junior contributors, companies begin questioning the need to expand the bottom of the pyramid.

So they simply don’t hire entry-level roles anymore, and teams become smaller and more experienced over time.

From a short-term efficiency perspective, this looks rational. From a long-term ecosystem perspective, however, it creates a lot of risk. As we discussed in a previous video, every senior engineer was once a junior, so if you are breaking the pipeline, the industry eventually runs out of experienced builders and technical leaders.

There is also an alternative few are discussing. Since AI could end up dramatically reducing the cost of building software, software might actually spread into industries that historically employed very few developers.

Healthcare systems could automate workflows, manufacturing could embed custom tooling, and agriculture could adopt data-driven automation. In that world, demand for developers increases rather than decreases, but the roles will look very different because entry-level developers will actually become domain specialists who understand both software and a specific industry.

But, whether we like it or not, the new reality is that junior developers cannot rely on time alone to create value.

The expectation shifts from learning slowly on the job to becoming productive quickly.

[2. Do skills matter?] Another important question affecting all of us is the importance of skills in a world where speed increases faster than understanding.

AI tools generate working code quickly, and developers adapt accordingly. If I’m being honest, whenever I’m using Claude or Codex, my instinctive response to a problem is to continue prompting rather than to think of a solution or go into the actual implementation. This changes how skills form.

A large majority of developers use AI assistance regularly now. Many newer

engineers encounter problems they never solve independently. They receive correct-looking solutions before developing the mental models that explain why those solutions work. This

is a big issue because AI-generated code often functions while hiding inefficiencies, security flaws, or architectural issues that only become visible under scale or stress.

A new study shows that 96% of engineers don’t fully trust AI output, yet only 48% verify it.

The report makes the contradiction very clear. Almost every engineer recognizes that AI output cannot be blindly trusted, yet less than half consistently verify what they ship. That gap between awareness and behavior is the real risk. And I think this

they ship. That gap between awareness and behavior is the real risk. And I think this gap will further increase. Because of tight deadlines or simply laziness, more and more developers will treat generated output as a finished solution instead of a starting point.

A dividing line emerges clearly. Developers who rely entirely on AI risk losing the ability to reason independently, while developers who use AI while maintaining strong fundamentals gain leverage.

[3. What is a software developer?!] But what’s more important is that the definition of a software developer is beginning to change. Coding itself is no longer the sole center of value creation so the role can evolve in two opposite directions.

In the pessimistic version, development becomes oversight. AI systems produce code, and developers review it for correctness, compliance, and safety. The creative aspect of programming diminishes, work becomes reactive rather than

and safety. The creative aspect of programming diminishes, work becomes reactive rather than generative and engineers spend more time validating outputs than designing solutions.

There are early signals of this transition. Engineers report increased time reviewing AI-generated pull requests, managing automated pipelines, and handling edge cases introduced by automated systems. Programming feels closer to quality assurance than creation.

A few days ago, a Tech crunch article on this topic made a lot of waves. According to Spotify, its best developers haven’t written a line of code since December, thanks to AI.

Of course, the post was met with a lot of skepticism on Reddit, but let’s not lie to ourselves, this is the direction all tech companies want to follow.

The alternative trajectory is more expansive. As AI systems take over implementation details, developers move upward in abstraction. They define system architecture, decide how components interact, and orchestrate multiple AI and software systems into coherent products.

[4. We should all be generalists] And this takes us to one of the most important shifts, in my opinion.

Rapid technological change exposes the weakness of narrow specialization. Historically,

deep expertise in a single technology could sustain an entire career. AI shortens the lifespan of such niches by reducing the effort required to operate within them.

Tasks that once justified specialization become automated quickly. UI building,

boilerplate API construction, or routine database optimization can increasingly be handled by tools.

Specialists who operate only at that layer risk seeing their value compressed. The industry

has seen this pattern repeatedly, from Flash developers to legacy frameworks. The difference

now is speed because AI accelerates how quickly a niche becomes commoditized.

What’s interesting is that at the same time, pure generalists struggle because shallow knowledge fails under complexity.

The emerging pattern favors T-shaped engineers, who are individuals with deep expertise in one or two areas combined with broad competence across adjacent domains.

These developers understand enough about multiple layers to make coherent decisions without becoming blocked by specialization boundaries.

[5. The real issue] But, if we take a step back, the real issue is that the legitimacy of traditional education is under pressure.

A four-year computer science degree historically served two purposes: teaching fundamentals and signaling competence to employers. AI challenges both. This is actually a more abstract topic, so we can tackle it in a separate video if you guys are interested in it.

We’ll change things a little bit this week, and instead of an awesome trivia, I have an awesome recommendation for you guys. A knight of the seven kingdoms is by far the best tv series I’ve seen in a long while, so if you want to escape the existential dread of AI for a couple of hours, this one is worth your time.

If you liked this video, you should consider joining our community where I’m posting more news and technical content. Please don’t forget to smash all the buttons, and, until next time, thank you for watching!