Software engineering is dead again

Every new model release, every new AI agent, every new benchmark, every new vibe coding tool, the same headline resurfaces: software engineering is dead. We have heard it ten thousand times now, and I am tired of it. The narrative is simple and seductive. AI writes code now, so developers are obsolete. Pay $200 a month for this tool and you will be safe. It is a classic marketing playbook: manufacture the fear, then sell the cure. And every time a new foundation model drops or a new coding agent launches, the cycle restarts. But the story does not hold up under scrutiny. Let me explain why.

The fear is manufactured

The "software engineering is dead" narrative has become a reliable content engine. Influencers, tool vendors, and even some CEOs lean into it because fear drives clicks and conversions. Salesforce's Marc Benioff said the company stopped hiring engineers. Headlines scream about mass obsolescence. The subtext is always the same: buy our product or get left behind. But here is the thing nobody selling you that story wants you to think about. If AI is truly capable of replacing software engineers, one of the most complex and demanding knowledge work roles that exist, then no job is safe. Not product managers, not designers, not marketers, not lawyers. Software engineering is where AI tooling lands first precisely because the work is so structured and well-defined compared to most other professions. Developers are the canary in the coal mine, not the only bird. So if you are panicking specifically about engineering roles, you are missing the bigger picture. Either AI replaces everything, or it replaces nothing fully. The middle ground, which is where reality actually sits, is far less dramatic than the headlines suggest.

What the data actually says

A CNN Business report from April 2026 put it plainly: the demise of software engineering jobs has been greatly exaggerated. Job openings for developers are actually growing, because companies now believe they can build more software than ever and need experienced engineers to shape those products. A widely discussed Reddit post cited MIT-affiliated research showing that of the 1.17 million tech workers laid off in 2025, only about 5% lost their jobs because AI directly automated their work. That is roughly 55,000 people out of over a million. The vast majority of layoffs were driven by post-COVID overcorrection, rising interest rates, and corporate restructuring, not by robots writing code. The Stanford AI Index 2026 report tells a more nuanced story. Employment among software developers aged 22 to 25 has declined nearly 20% since 2024, while headcount for developers aged 35 to 49 has grown by 9%. Hiring is happening, just not evenly. Entry-level roles are getting squeezed, but experienced engineers are more in demand than ever. Gartner predicts that by 2027, generative AI will prompt 80% of engineers to upskill, not because their jobs are disappearing, but because the nature of the work is shifting. The role is evolving from writing boilerplate to orchestrating systems, reviewing AI-generated output, and making architectural decisions that no model can reliably make on its own.

AI is still bad at the hard parts

Here is what the hype merchants conveniently leave out: AI is still remarkably poor at generalizing across tasks. It can autocomplete functions, generate boilerplate, and scaffold projects. But the moment you need it to reason about a complex distributed system, understand subtle business requirements, debug a production incident at 2 AM, or make a judgment call about technical debt, it falls apart. A Stack Overflow developer survey found that 46% of developers actively distrust the accuracy of AI-generated code, while only 3.1% express high trust. The most experienced developers, the ones best positioned to evaluate quality, are the most skeptical. There is a reason for that. They have seen what happens when you ship code that "looks right" but is not. Every AI coding tool still needs a human behind it. Someone has to review, test, deploy, monitor, and take responsibility. The idea that you can replace an engineering team with a subscription is, to put it mildly, not supported by reality.

The layoffs are not what they seem

The New York Post reported that 52,050 tech jobs were cut in just the first quarter of 2026, a 40% jump from the same period last year. AI was increasingly cited as the reason. But a Guardian investigation from the same month painted a different picture: companies like Microsoft, Amazon, and Block were making cuts driven by years of overhiring during the pandemic boom, not because AI had suddenly automated those roles away. A majority of CEOs surveyed reported no financial returns from their AI investments yet. The layoffs and the AI spending are happening simultaneously, but correlation is not causation. Companies are using "AI transformation" as a convenient narrative to justify cost-cutting that was coming regardless. As one commenter in a widely shared Reddit thread put it: "They weren't letting those people go because of AI. They were let go because our economy is in freefall. AI innovation was the more palatable story to sell stakeholders and the public."

The real problem: interviews are stuck in 2010

Here is what actually frustrates me. We have moved so aggressively toward AI-assisted development, where engineers use copilots and agents daily, but the hiring process has not caught up at all. Companies are still asking candidates to solve LeetCode problems on a whiteboard. Invert a binary tree. Implement a linked list. Memorize dynamic programming patterns. These problems have almost nothing to do with actual software engineering. They never really did, and the gap between interview and reality has only gotten wider. If your engineers are using AI tools every day on the job, why are you testing their ability to solve algorithmic puzzles from memory in a 45-minute pressure cooker? The good news is that some companies are waking up. HackerEarth reported that aptitude-based assessments surged 54x since 2024, as companies shift toward testing problem-solving ability and judgment over raw syntax recall. Take-home projects and work simulations are gaining ground over traditional coding interviews. It is long overdue. As the Pragmatic Engineer newsletter noted, the industry is shifting from "how" to build to "what" to build. The engineers who thrive will be the ones who can define problems clearly, make sound architectural decisions, and orchestrate AI tools effectively, not the ones who memorized the most algorithms.

What is actually happening

The role of the software engineer is changing, not dying. Developers are becoming orchestrators rather than typists. The tedious parts of the job, writing boilerplate, scaffolding CRUD endpoints, generating test stubs, are increasingly handled by AI. What remains, and what is growing in importance, is everything else: system design, debugging complex interactions, understanding user needs, making tradeoffs, and taking ownership of outcomes. Teams that once had ten junior developers are now operating with two senior engineers supported by AI agents. That is a real structural shift, and it does mean fewer entry-level positions in the short term. But it also means the ceiling for what a small team can build has never been higher. More software will be built, not less. And someone still needs to build it well.

Stop buying the fear

The next time someone tells you software engineering is dead, ask yourself: who benefits from you believing that? Usually it is someone selling a course, a tool, or a subscription. The narrative is designed to create urgency and anxiety so you open your wallet. Software engineering is not dead. It is changing, like it always has. The engineers who adapt, who learn to work with AI tools instead of competing against them, who focus on judgment and architecture over syntax, will be fine. The ones who panic and stop learning will struggle, but that has always been true in this industry, AI or not. So take a breath. Close the doom-scrolling tab. And go build something.