There has been a lot of talk about the promises of generative AI tools and how they are improving the lives of software developers in recent years. Recently, things have been heating up with new tools like GitHub’s Copilot touting massive benefits to productivity for developers and profitability for management. For some, it has even brought the software developer profession itself into question!

Despite this tool's hype, the developer community has mixed opinions on its role in the industry and their own personal workflows. While Copilot does offer useful suggestions and can streamline simple coding tasks, its true utility remains a bit foggy.

Some believe it is the future of programming, while others warn young developers of the risks of becoming dependent on its assistance.

Thinking about my experience as a developer over the past 20 or so years, I’ve heard it all. The promise of greater efficiency due to automation and AI. The redundancy of developers as new technologies learn to do our jobs for us. These are ideas I’ve heard again and again.

Around 20 years ago now, platforms like WordPress and various template-based systems came along. Various other tools also came during these early days that made coding easier, like JetBrains and other IDEs. These tools increased the efficiency of the developer and helped get a product out to market faster and more efficiently.

Building websites used to be a real grind. You had to know your way around HTML, CSS, and JavaScript - it really did take forever.

Then, WordPress came along and changed the game. With pre-built, easy-to-use tools, anyone could start putting up a website. It leveled the playing field for everyone.

Sure, WordPress made things a lot easier, but it didn't solve every problem.

You still needed serious coding chops for complex sites and custom use cases. Plus, security, speed, and scaling were always headaches.

That's where the next big wave of tools came in. Ruby on Rails, Django, and Node.js were like turbochargers. They allowed us to build things much faster and with cleaner code. It was a game-changer.

While these advancements undoubtedly improved developer productivity, they also brought new complexities. As the software ecosystem grew, developers faced the challenge of staying up-to-date with new languages and use cases, which made it important to upskill to remain competitive. The demand for specialized skills, such as frontend and backend development, enterprise-level stability, and scalability all increased.

Instead of being responsible for every aspect of the code, the software developer role was evolving to be more focused on higher-level tasks and strategy rather than some of the simple operations that were common in the early days.

As it turns out, these trends were foreshadowing the effect that generative AI tools like Copilot were going to have on the somewhat idealistic developer community leading up to 2020.

The Promises and Initial Impact of Code Generators

The advent of AI-powered coding assistants like GitHub Copilot ushered in a wave of optimism that had developers and management alike hopeful for the improvements it promised to bring. In 2021 and 2022, these tools were positioned as game-changers, promising to revolutionize the software development lifecycle.

Sam Altman touted the tool as a game-changer back in 2021. A pretty reliable source, considering what was to come from OpenAI:

Others mocked the tool as merely a new version of Clippy (yeah, that tool that Microsoft created as an assistant in Office tools!)

There was a dizzy phase of excitement (and fear!) at the sight of code being generated automatically, (sort of like the excitement of seeing words appear with the wider launch of ChatGPT  - but that came later!) Even influencers such as @Cassidoo explained how while some of the features were “hilarious” and only “partially right”, using CopIlot was overall an “impressive experience”.

Others were mocking its capabilities at the time, with some having harsh opinions such as concluding that CoPilot isn’t going to affect the industry at all.

Some were even questioning its legitimacy and accusing it of using code from open-source libraries:

At the time, it was obvious that there was something there, but it was too early to tell what impact it was going to have and how useful it really was.

It was obvious it could help with boilerplate coding tasks and general efficiency, but GitHub themselves were transparent in talking about the many improvements that were to come over the next few years.

As I researched the topic further, there were clear efficiency improvements from Copilot, touted as 20-30% in many cases, although I have not seen much apart from streamlining simple tasks and offering creative suggestions.

Don't get me wrong, a 30% increase in efficiency is no small matter. GitHub estimates there to be a $1.5T saving across the industry. Sounds a bit aspirational, albeit a little dubious to me (especially considering the source of these stats comes from the creators themselves). Their article does reference a study that mentions productivity gains of up to 55%, so this does give us some pause for thought.

While plenty of GitHub articles are touting these and other monumental stats such as “92% of U.S.-based developers are already using AI coding tools both in and outside of work”, it's important to find some outside opinions on this matter as well.

Similar to the developers mentioned above, some leaders in the space are not so in love with Copilot in 2024. The co-creator of Django called Copilot a “fancy autocomplete.”

So with all of these opinions erupting after the introduction of such a divisive tool, where have things progressed in the past few years?

As we will see, these code generators had a bit of a different impact than just increasing efficiency and productivity. Before we get into that, let’s check out what people are saying about Copilot in 2024.

The Effects of Coding Assistants on Developer's Workflows: Riding the Waves of Change

In just a few short years, alot had changed!

Starting in late 2022 and into 2024, things had begun to take off for Copilot (no pun intended).

GitHub execs were aiming to make developers 10x more productive, and the company’s CEO Thomas Dohmke claimed that soon Copilot will be writing 80% of code.

The speed, efficiency, and utility of these tools have made big leaps since the early 2020s.

In a 2022 study performed internally by Github, they split a group of 95 developers into two groups, one with Copilot, and one without. Then they asked them to write a web server in JavaScript. The group with Copilot was not only more likely to finish the task (78% to 70%), but they did so 55% faster.

The study also found that 73% of Copilot users felt “more in the flow,” while 74% reported being able to “focus on more satisfying work.”

While these numbers are impressive, let’s take a look at some outside sources to see if people who are not involved with GitHub feel the same way:

In 2023, leading developers with deep AI training and expertise on the subject such as @HamelHusain highlighted key improvements such as:

• Added ability to plan and write entire projects
• Improved ability to fix bugs
• Write complex terminal commands

Other leading senior engineers like Sachidanand Sharma wrote great articles detailing its most prominent ups and downs:

“The most immediate benefit of Copilot is the undeniable boost it provides to my coding speed. Repetitive tasks like writing boilerplate code, generating getters and setters, and implementing common data structures are now handled effortlessly by Copilot's suggestions. This frees up valuable mental space and allows me to focus on the more challenging and creative aspects of programming.”

But despite the time-saving advantages and increased creativity and output, Sachidanand forewarned of it not being a silver bullet.

“Despite its remarkable capabilities, it's important to acknowledge that GitHub Copilot is not a silver bullet. The suggested code requires review and understanding before blindly implementing it. Additionally, it can sometimes generate incorrect or suboptimal solutions, so a critical eye is always necessary.”

As it turns out, this is a critical point that is representative of the market at large currently.

While Copilot and similar tools like Cursor do offer benefits to speed and productivity, many leading developers are warning that these upgrades may only apply to experienced developers with the in-depth knowledge and experience required to understand the nuanced context surrounding the code that is generated.

As Sachidanand puts it, “Context is king. Don't treat Copilot as a code generation machine in isolation.”

While the industry at large seems to agree that the productivity and efficiency benefits are nothing to overlook, these advancements have come with some important caveats. These include a need to understand the context surrounding the code generated, and an understanding of the deeper strategy and nuances behind the “why” of the code.

Users seem to agree that even advanced and hyped-up models like Devin can’t handle deeper tasks, and need teams who understand strategy, proper implementation, and context.

Leading tech influencer Clement Mihailescu recently provided his take on these effects on the current market with the following:

Here is another take from Sidi Jeddou, a developer from Rapid Forms who talks about the “10X” developer” idea that has been bouncing around the space for a while now:

Additionally, some of the developer community are concerned that these tools may only be able to be properly utilized by experienced and thoughtful developers who understand how to design prompts to maximize the chance of a successful output:

So what does this mean for entry-level developers? And what does this mean for the state of the industry as a whole?

The Future of Generative AI Tools for Developers and The New “Senior Developer Market”

Putting all of this together, we are seeing two main themes:

1. A changing job market

It is more competitive than ever, with thousands of junior developers struggling to find entry-level positions, as well as mid-senior-level developers with 5+ years of experience dominating the job market.

2. A shift in skills

A clear shift in the demand for skills in software engineers, such as prompt engineering, strategic thinking, and complex systems design.

These changes are leading to a tangible reaction from many junior developers in the space, even those who are confident in their skills and are leaders of the new generation of developers.

According to influencer Namanh Kapur in his recent video titled “The Broken Senior-Only Developer Market”,  his recent job search alongside various other entry-level developers has found that the current tech job market heavily favors mid to senior-level developers who have at least 5+ years of experience.

Namanh and his friends spent weeks searching through job boards, and had a very difficult time finding any listings that were friendly to developers with less than 3-5 years of experience.

While this new market may be jarring to new developers who were hopeful to take part in the gold rush they had heard so much about a decade ago, there is still optimism among those who understand the need for adaptation:

It may be helpful now to look at the hard facts about what actions leading companies are taking in regard to their workforce. Are less experienced developers really getting laid off en masse?

According to payroll giant ADP, the availability of software developer jobs has experienced a steady decline since 2019, with the authors stating: “The emergence of artificial intelligence might be a reason for the shift, as employers invest in automation.” This echoes many of the fears and warnings of the community we’ve seen so far.

According to a recent LA Times article, “AI A Job Killer? In California, It’s Complicated”, there were massive layoffs from prominent companies such as Block, Google, Toast, and Meta. However, these layoffs were not strictly due to superhero developers using AI.

In fact, they were mainly due to “stricter investor demands — what managers saw as over-hiring during the pandemic and a stock market that rewarded those personnel cuts.”

They also note how AI still did play a role in these layoffs, saying ”...more than 28,000 job cuts by tech firms nationwide were announced in the first two months of this year…only a few hundred of those layoffs were explicitly attributed to AI, but thousands more cuts in tech and other industries were said to be the result of “updating or incorporating new technology”.

While this language does cause some pause for concern, experts also agree that these layoffs are a direct result of the industry bouncing back into equilibrium after excessive pre-pandemic hiring.

Experts in the article also discuss how these tools are having a deep impact on the way organizations are being run, and integrating AI into the business model is now a key business goal for a majority of these players. They note the need for developers to learn new skills that fill the new roles that are soon to be created.

Charles Lee Isbell Jr., who studied at MIT’s AI Lab, mentions in the article that “It’s not a death knell.”

Building off of this, Ayanna Howard, a robotics expert and dean of the engineering college at Ohio State University, leaves us with the following:

“AI may ultimately lead to a smaller tech workforce. But, there will also be AI-augmented jobs, as well as new jobs. One position in high demand is prompt engineer, who designs prompts and other processes to get optimal performance from generative AI tools, whether text or images.”

“We don’t necessarily know what it is that we need,” “But they will come out, and when they do, it’ll be like, ‘Oh, guess what? It’s another field.”

To wrap this all up, the software engineer market is just not what it was in the early days. After a gold rush-esque period, the industry has done a sort of course-correct recently. Key differences now include factors such as a higher barrier to entry, more competition for junior developers, and a need to adapt to new organizational demands.

While the field may be dominated by mid to senior-level developers, there is room for upskilling, deep learning, and even for new job titles, such as prompt engineers, systems theory experts, machine learning developers, and complexity and Agile strategists with development expertise.

This is not a death sentence for new developers, or the developer market, though!

As we saw in the first wave of changes with the advent of WordPress early on, the main effect here is a need for adaptation to the new state of the industry, and even the need for programming by society at large.

When we zoom out, we can see that the developer has yet to become a dying breed. However, junior developers may need to consider what types of skills are helping their mid-senior-level counterparts enjoy success and what types of skills and use cases leading organizations have their eyes on.

Thanks for reading, and would love to hear your thoughts on this!

-Yossi