What is the story about?
The debate around AI replacing jobs does not seem to end. Amid the rising anxiety, we sought to bring you a clarity over the issue. In this series, we will try to target every profession that is most vulnerable to AI. The first article targets the software engineers aka the coders.
Not long ago, the journey into coding followed a familiar script. You started small, fixed bugs, wrote basic functions, and slowly climbed your way up the ladder. Today, that ladder looks different. Some of its lowest rungs may already be disappearing.
The recent research by Anthropic listed the top 10 professions most exposed to AI disruption, and coding took the top spot. And it is not just theoretical. The anxiety is already visible with every wave of AI-driven layoffs.
A telling example of this shift comes from Andrej Karpathy, who has witnessed the transformation firsthand. The former OpenAI cofounder and ex-head of AI at Tesla recently revealed that he has barely written code manually in months, a change he admits feels both remarkable and unsettling.
Speaking on the No Priors podcast, Karpathy described how his workflow flipped almost overnight around December. Where he once wrote the bulk of his code himself, he now delegates nearly all of it to AI systems, stepping in primarily to guide and refine the output. What used to be hours or even days of hands-on coding has, in many cases, been compressed into minutes with the help of AI.
The shift, he suggested, is not gradual but deeply abrupt. Tasks that once defined a developer’s day are increasingly being handled by machines, leaving humans in more supervisory roles.
For Karpathy, the experience has been both efficient and disorienting, a glimpse into just how quickly the foundations of software development are being rewritten.
On a similar note, Sam Altman took to X recently, reflecting on how far software development has come, paying tribute to programmers who once built complex systems line by line.
He noted that it is already becoming difficult to recall the sheer effort that coding demanded in the past, thanking developers for laying the groundwork that today’s AI-driven tools are building upon.
His remarks arrive at a time when AI can generate code in seconds, fix bugs almost instantly, and even recommend cleaner, more efficient structures. Tasks that once required hours of focused work can now be completed in a fraction of the time, fuelling fresh anxiety about the future of coding jobs, particularly at the entry level.
Yet, one question continues to linger: is AI making life easier for entry-level developers, or quietly pushing them out of the job market?
Let’s start with the uncomfortable truth. A large portion of entry-level coding work involves repetitive, well-defined tasks: writing boilerplate code, debugging simple errors, documenting functions, or building basic components. These are exactly the areas where AI excels.
Modern AI tools like Anthropic Claude, OpenAI Codex, Google Gemini, and Perplexity are no longer just assistants, they are increasingly capable collaborators. From writing and reviewing code to optimising performance, these systems are proving their mettle across developer workflows.
Benchmarks such as SWE-Bench show that advanced models can now tackle real-world GitHub issues with growing accuracy, at times coming close to human-level performance.
What makes this shift particularly significant is not just speed, but scope. AI is increasingly handling the kind of repetitive, structured tasks that traditionally defined entry-level roles, allowing smaller teams to build and ship products faster than ever.
As a result, coding is no longer just about writing lines of code from scratch, but about directing, reviewing, and refining outputs generated by intelligent systems.
In enterprise settings, Claude in particular has found a strong foothold in reasoning-heavy and coding-centric tasks. Engineering teams are using it for everything from code generation and debugging to explaining system architecture and producing technical documentation.
The momentum is also reflected in market trends. According to Menlo Ventures, Anthropic captured an estimated 54 per cent share of the coding-focused AI market by December 2025, up from 42 per cent just six months earlier, largely driven by the adoption of Claude Code.
At the same time, OpenAI continues to dominate the broader chatbot landscape, with ChatGPT holding an 80.04 per cent market share as of February 2026, while Claude accounts for 1.34 per cent, according to StatCounter.
Reports have highlighted how companies are already relying on AI tools to automate tasks that were once assigned to junior engineers. Startups, in particular, are able to operate with smaller teams, using AI as a “force multiplier” to reduce hiring costs.
The economic logic is hard to ignore. If a company can use an AI tool to generate functional code in seconds, at a fraction of the cost of hiring a junior developer, the incentive to cut entry-level roles becomes very real.
Moreover, a growing number of studies and industry observations are raising red flags about the skill levels of recent graduates entering the workforce. Many new computer science degree holders appear to depend heavily on AI tools to complete coding tasks, often without fully grasping the underlying concepts.
This overreliance has sparked concerns that while they can produce working outputs, their understanding of core principles and the systems they build may be shallow.
However, the story does not end there. For all its strengths, AI remains far from perfect.
One of the biggest challenges is hallucination, where AI generates incorrect or misleading outputs with confidence. In coding, this can mean subtle bugs, inefficient logic, or even security vulnerabilities that are not immediately obvious.
Even the most advanced models require oversight. They can assist, accelerate, and augment, but they still depend on human judgement for validation. Senior developers are often needed to review AI-generated code, ensure it aligns with project requirements, and integrate it into larger systems.
There is also the question of context. AI may generate correct code snippets, but understanding the broader architecture of a product, user needs, and long-term scalability still requires human expertise.
In that sense, AI is less of a replacement and more of a collaborator, albeit a very fast and increasingly capable one.
The debate has also reached industry leaders, many of whom are urging caution before writing off entry-level roles entirely.
Zoho co-founder Sridhar Vembu has raised a critical concern: if AI eliminates entry-level jobs, where will future senior engineers come from? After all, every experienced developer starts somewhere.
His argument highlights a structural risk. Entry-level roles are not just about productivity; they are training grounds. They allow individuals to learn, experiment, and gradually build the skills needed for more complex responsibilities.
If these opportunities shrink significantly, companies may face a talent gap in the long term. The pipeline of skilled engineers could weaken, creating challenges that AI alone cannot solve.
Based on current trends, it is fair to say that AI is already reshaping the landscape. Certain roles, especially those centred on repetitive tasks, may become redundant overtime. The traditional definition of a “junior developer” could become obsolete.
But that does not necessarily mean the end of entry-level opportunities.
Instead, these roles may evolve. Future entry-level developers might be expected to work alongside AI, focusing less on writing code from scratch and more on guiding, testing, and refining AI outputs. Skills like prompt engineering, system thinking, and code review could become more valuable than ever.
In other words, the bar may rise, but the door may not fully close.
The reality, as always with technology, lies somewhere in between. AI is not just replacing jobs; it is redefining them. For aspiring developers, the challenge will be to adapt quickly and learn how to work with AI, not against it.
The question is no longer whether AI will change coding jobs. It already has. The real question is who will adapt fast enough to stay relevant.
Not long ago, the journey into coding followed a familiar script. You started small, fixed bugs, wrote basic functions, and slowly climbed your way up the ladder. Today, that ladder looks different. Some of its lowest rungs may already be disappearing.
The recent research by Anthropic listed the top 10 professions most exposed to AI disruption, and coding took the top spot. And it is not just theoretical. The anxiety is already visible with every wave of AI-driven layoffs.
Tech leaders agree
A telling example of this shift comes from Andrej Karpathy, who has witnessed the transformation firsthand. The former OpenAI cofounder and ex-head of AI at Tesla recently revealed that he has barely written code manually in months, a change he admits feels both remarkable and unsettling.
Speaking on the No Priors podcast, Karpathy described how his workflow flipped almost overnight around December. Where he once wrote the bulk of his code himself, he now delegates nearly all of it to AI systems, stepping in primarily to guide and refine the output. What used to be hours or even days of hands-on coding has, in many cases, been compressed into minutes with the help of AI.
The shift, he suggested, is not gradual but deeply abrupt. Tasks that once defined a developer’s day are increasingly being handled by machines, leaving humans in more supervisory roles.
For Karpathy, the experience has been both efficient and disorienting, a glimpse into just how quickly the foundations of software development are being rewritten.
On a similar note, Sam Altman took to X recently, reflecting on how far software development has come, paying tribute to programmers who once built complex systems line by line.
He noted that it is already becoming difficult to recall the sheer effort that coding demanded in the past, thanking developers for laying the groundwork that today’s AI-driven tools are building upon.
I have so much gratitude to people who wrote extremely complex software character-by-character. It already feels difficult to remember how much effort it really took.
Thank you for getting us to this point.
— Sam Altman (@sama) March 17, 2026
His remarks arrive at a time when AI can generate code in seconds, fix bugs almost instantly, and even recommend cleaner, more efficient structures. Tasks that once required hours of focused work can now be completed in a fraction of the time, fuelling fresh anxiety about the future of coding jobs, particularly at the entry level.
Yet, one question continues to linger: is AI making life easier for entry-level developers, or quietly pushing them out of the job market?
AI is already doing the “junior-level” work
Let’s start with the uncomfortable truth. A large portion of entry-level coding work involves repetitive, well-defined tasks: writing boilerplate code, debugging simple errors, documenting functions, or building basic components. These are exactly the areas where AI excels.
Modern AI tools like Anthropic Claude, OpenAI Codex, Google Gemini, and Perplexity are no longer just assistants, they are increasingly capable collaborators. From writing and reviewing code to optimising performance, these systems are proving their mettle across developer workflows.
Benchmarks such as SWE-Bench show that advanced models can now tackle real-world GitHub issues with growing accuracy, at times coming close to human-level performance.
What makes this shift particularly significant is not just speed, but scope. AI is increasingly handling the kind of repetitive, structured tasks that traditionally defined entry-level roles, allowing smaller teams to build and ship products faster than ever.
As a result, coding is no longer just about writing lines of code from scratch, but about directing, reviewing, and refining outputs generated by intelligent systems.
In enterprise settings, Claude in particular has found a strong foothold in reasoning-heavy and coding-centric tasks. Engineering teams are using it for everything from code generation and debugging to explaining system architecture and producing technical documentation.
The momentum is also reflected in market trends. According to Menlo Ventures, Anthropic captured an estimated 54 per cent share of the coding-focused AI market by December 2025, up from 42 per cent just six months earlier, largely driven by the adoption of Claude Code.
At the same time, OpenAI continues to dominate the broader chatbot landscape, with ChatGPT holding an 80.04 per cent market share as of February 2026, while Claude accounts for 1.34 per cent, according to StatCounter.
Reports have highlighted how companies are already relying on AI tools to automate tasks that were once assigned to junior engineers. Startups, in particular, are able to operate with smaller teams, using AI as a “force multiplier” to reduce hiring costs.
The economic logic is hard to ignore. If a company can use an AI tool to generate functional code in seconds, at a fraction of the cost of hiring a junior developer, the incentive to cut entry-level roles becomes very real.
Moreover, a growing number of studies and industry observations are raising red flags about the skill levels of recent graduates entering the workforce. Many new computer science degree holders appear to depend heavily on AI tools to complete coding tasks, often without fully grasping the underlying concepts.
This overreliance has sparked concerns that while they can produce working outputs, their understanding of core principles and the systems they build may be shallow.
But AI still needs a human in the loop
However, the story does not end there. For all its strengths, AI remains far from perfect.
One of the biggest challenges is hallucination, where AI generates incorrect or misleading outputs with confidence. In coding, this can mean subtle bugs, inefficient logic, or even security vulnerabilities that are not immediately obvious.
Even the most advanced models require oversight. They can assist, accelerate, and augment, but they still depend on human judgement for validation. Senior developers are often needed to review AI-generated code, ensure it aligns with project requirements, and integrate it into larger systems.
There is also the question of context. AI may generate correct code snippets, but understanding the broader architecture of a product, user needs, and long-term scalability still requires human expertise.
In that sense, AI is less of a replacement and more of a collaborator, albeit a very fast and increasingly capable one.
Industry leaders warn
The debate has also reached industry leaders, many of whom are urging caution before writing off entry-level roles entirely.
Zoho co-founder Sridhar Vembu has raised a critical concern: if AI eliminates entry-level jobs, where will future senior engineers come from? After all, every experienced developer starts somewhere.
His argument highlights a structural risk. Entry-level roles are not just about productivity; they are training grounds. They allow individuals to learn, experiment, and gradually build the skills needed for more complex responsibilities.
If these opportunities shrink significantly, companies may face a talent gap in the long term. The pipeline of skilled engineers could weaken, creating challenges that AI alone cannot solve.
So, are entry-level coding jobs truly at risk?
Based on current trends, it is fair to say that AI is already reshaping the landscape. Certain roles, especially those centred on repetitive tasks, may become redundant overtime. The traditional definition of a “junior developer” could become obsolete.
But that does not necessarily mean the end of entry-level opportunities.
Instead, these roles may evolve. Future entry-level developers might be expected to work alongside AI, focusing less on writing code from scratch and more on guiding, testing, and refining AI outputs. Skills like prompt engineering, system thinking, and code review could become more valuable than ever.
In other words, the bar may rise, but the door may not fully close.
The reality, as always with technology, lies somewhere in between. AI is not just replacing jobs; it is redefining them. For aspiring developers, the challenge will be to adapt quickly and learn how to work with AI, not against it.
The question is no longer whether AI will change coding jobs. It already has. The real question is who will adapt fast enough to stay relevant.
