|
Listen to this story
|
“Programming is no longer hard” or “everyone’s a developer” are the most common phrases that one would hear on LinkedIn or X as everyone is basically talking about Cursor, Claude, or GitHub Copilot. But the problem is that most of the people who claim so are not developers themselves. They are merely ‘modern developers.’
Register for NVIDIA AI Summit India
Santiago Valdarrama, founder of Tideily and an ML teacher, who has been actively asking developers if they are using Cursor or not, started another discussion that tools such as Cursor and others are basically tools that can only assist existing developers in writing better code. “Wake me up when companies start hiring these clueless modern ‘developers’,” he added.
He gave an analogy of calling yourself an F1 racer after playing a racing game on an iPad.
In all honesty, it is undeniable that the barrier to entry for becoming a developer has dropped significantly ever since Cursor, even ChatGPT, dropped. People have been able to build software for their personal use and even build apps in mere hours. But, this does not eliminate the fact that it is currently only limited to just creating such apps and low level software.
“You Can’t Correct Code if You Don’t Know How to Code”
Given all this hype around the end of software engineering roles, developers and programmers are getting worried about the future of their jobs. It is indeed true that software engineers have to upskill faster than anyone else, but the fear of getting replaced can be pushed off to at least a few years.
Having tools such as Cursor and Claude are only good enough if a developer actually knows how the code actually works. The real game-changer is how developers who use AI will outpace those who don’t. “The right tools can turn a good developer into a great one. It’s not about replacing talent; it’s about enhancing it,” said Eswar Bhageerath, SWE at Microsoft.
AI only takes care of the easy part for a developer – writing code. The real skill that a developer has is reasoning and problem solving, apart from fixing the bugs in the code itself, which cannot be replaced by any AI tool, at least anytime soon. Cursor can only speed up the process and write the code but correcting the code is something that only developers can do.
Moreover, bugs generated within code with AI tools are not easily traceable by developers without using any other AI bug detection tool. Andrej Karpathy, who has been actively supporting Cursor AI over GitHub Copilot, also shared a similar thing while working. “it’s slightly too convenient to just have it do things and move on when it seems to work.” This has also led to the introduction of a few bugs when he is coding too fast and tapping through big chunks of code.
These bugs cannot be fixed by modern ‘developers’ who were famously also called ‘prompt engineers’. To put it simply, someone has to code the code for no-code software.
Speaking of prompt engineers, the future will include a lot of AI agents that would be able to write code themselves. The future jobs of software engineers would be managing a team of these AI coding agents, which is not possible for developers who just got into the field by just learning to build apps on Cursor or Claude. It is possible that the size of the teams might decrease soon as there would be no need for low level developers.
Upskilling is the Need of the Hour
That is why existing developers should focus on developing engineering skills, and not just coding skills. Eric Gregori, adjunct professor at Southern New Hampshire University, said that this is why he has been teaching his students to focus more on engineering than just programming. “AI is too powerful of a tool to ignore,” he said, while adding that existing limitations of coding platforms have been removed completely.
“Hopefully, AI will allow software engineers to spend more time engineering and less time programming.” It is time to bring back the old way of learning how to code as modern developers would be tempted to just copy and paste code from AI tools, and not do the real thinking.
The F1 driver analogy fits perfectly here. Most people can learn how to drive, but would never be able to become a race driver. The same is the case with the coding tools. But if all people need is prototyping and designing an initial code, AI driven developers would be able to do a decent enough job.
That is why a lot of pioneers of the AI field such as Karpathy, Yann LeCun, Francois Chollet, even Sam Altman, say that there would be 10 million coding jobs in the future, the ones that would require the skills of Python, C++, and others. As everyone in some way would be a ‘modern developer’, and most of the coding would be done by AI agents.
It is possible that most of the coding in the future would be in English, but most of it would be about debugging and managing the code generated by AI, which is not possible for someone who does not know coding from scratch.
