Can AI replace programmers?

Last updated: July 2026

AI can now take a plain-English request and write working code in seconds. For a job that's literally about writing code, that sounds like an existential threat. But look closely at what programmers actually do all day, and the picture gets more interesting than "the robots are taking over."

The question comes up constantly, and it's a fair one. If an AI can write code, and a programmer's job is to write code, surely programmers are first in line to be automated? Some prominent voices in tech have leaned into exactly this. Boris Cherny, the creator of Anthropic's Claude Code, said in early 2026 that "coding is practically solved" and predicted the job title "software engineer" might give way to just "builder".¹

And yet, so far, the mass replacement hasn't happened. To understand why, you have to look past the headline and ask what the job actually involves.

Writing code was never the bottleneck

Here's the fact that reframes the whole debate. Multiple studies have found that programmers spend surprisingly little of their time actually writing code, somewhere between 9% and 61% depending on the study, with one large analysis of 6,000 developers at Microsoft landing in that range.² The rest of the job is understanding what needs building, designing how the pieces fit, testing, debugging, reviewing other people's work, and talking to the people who'll use the software.

This matters enormously. If typing out code were the main job, automating it would gut the profession. But if coding is a minority of the work, then a tool that speeds up coding makes developers faster without making them unnecessary. As one analysis noted after a wave of companies adopted AI coding agents in late 2025, using agents to write most of the code led to surprisingly little impact on overall productivity, because writing code was never the real constraint.³

"AI won't replace programmers; programmers using AI will replace programmers not using AI."

That line has become a cliché in the industry, but it captures the current reality well. The developers who thrive are the ones using AI as a fast assistant, not the ones being replaced by it.⁴

What AI is genuinely good at (and not)

AI coding tools are legitimately impressive at certain things. Controlled experiments have shown they can speed up well-defined coding tasks significantly; one widely cited GitHub study found around 55% faster completion on a specific task using an AI assistant.⁴ They're strong at boilerplate, standard functions, simple bug fixes, writing tests, and explaining unfamiliar code.

Where they struggle is exactly where senior developers earn their keep: work involving ambiguity, complex systems with many interdependencies, and real-world constraints that aren't written down anywhere.⁵ An AI can write a function if you specify it precisely. It's much weaker at deciding what to build, spotting that a requirement doesn't make sense, or untangling a subtle bug that spans five different systems. Those tasks need judgement built from experience, and they're the heart of the job.

It's also worth remembering, as we've written elsewhere, that AI models confidently make things up. Code that looks right but is subtly wrong is arguably more dangerous than code that obviously fails, and catching that still needs a human who understands the system.

The real shift: a squeeze on junior roles

The honest, uncomfortable part of the story is about entry-level work. The tasks AI handles best, simple, well-defined coding, are exactly the tasks junior developers used to cut their teeth on. Several analyses describe a job market that looks contradictory: fewer junior openings, but strong and growing demand for senior engineers who can direct AI tools and review their output.⁶

One study observed a 21% drop in coding-related job postings in the period after ChatGPT's introduction, though the same research stressed this reflects changing demand rather than wholesale elimination.⁷ The pattern is less "replacement" and more "rebalancing": the industry is shifting its weight from junior to senior, from writing code to reviewing and directing it. That's a real and difficult change for people trying to enter the field, even if the total number of jobs holds up.

What the job numbers actually say

Despite the disruption, the aggregate forecasts remain surprisingly healthy. The US Bureau of Labor Statistics projects software developer employment to grow, with some analyses citing around 17% growth through 2033, adding hundreds of thousands of roles.⁶ A McKinsey analysis frequently cited in this debate estimates that as much as 80% of programming work will still require human input even as AI automates more tasks.⁸

History backs this up in a way that's easy to forget. Every previous tool that made coding dramatically easier, compilers, high-level languages, Stack Overflow, cloud platforms, was predicted to reduce the need for programmers. Each time, the opposite happened: easier coding led to more software being built, which led to more programming jobs, not fewer.¹ Whether AI breaks that pattern or repeats it is the genuine open question.

So, can AI replace programmers?

The most accurate answer is: it's replacing parts of the job, not the job. AI is very good at the coding tasks that were never the hard part, and much weaker at the design, judgement, and system-level thinking that were. The result isn't a profession being wiped out. It's a profession being reshaped, with the work moving up the ladder toward design and review, and real pressure landing on the entry-level rungs.

For anyone in or entering the field, the practical implication is consistent across nearly every source: learn to use these tools well, and lean into the parts of the work AI can't do, understanding problems, designing systems, and exercising judgement. The programmers most at risk aren't the ones competing with AI. They're the ones who could have used it and didn't.

Sources

1. Frontend Highlights, "Will AI Replace Programmers in 2026–2027?" (2026), including Boris Cherny remarks — medium.com
2. NormalTech, "Why AI hasn't replaced software engineers, and won't" (2026), citing a 2019 summary and Microsoft data — normaltech.ai
3. Ibid. (on coding agents and productivity, late 2025).
4. Sanjay Negi, "No — 4 Out of 5 Developers Are Not Losing Their Jobs to AI" (2025), citing McKinsey and a GitHub Copilot experiment — medium.com
5. Analytics Insight, "Will AI Replace Programmers and Developers?" (2026) — analyticsinsight.net
6. Index.dev, "Will AI Replace Developers? 2026 Job Market Reality," citing the 2025 Stack Overflow survey and BLS projections — index.dev
7. Demirci et al., cited in "LLMs' Reshaping of People, Processes, Products, and Society in Software Development," arXiv:2503.05012 (2025) — arxiv.org
8. Relevant Software, "Will AI Replace Programmers?" (2026), citing McKinsey — relevant.software