In March of this year, artificial intelligence led all reasons for job cuts in the United States, accounting for 15,341 of the month’s 60,620 announced layoffs. That’s 25% of all layoffs announced in March, up from about 10% in February. Since companies started listing AI as a cause of layoffs in 2023, it has appeared in close to 100,000 layoff notices.
Taken at face value, the narrative is simple: AI is causing job losses. Dig a little deeper, however, and you’ll find something else.
AI is not simply cutting roles. It is fundamentally rewriting what those roles require, how companies evaluate talent, and which skills separate employable engineers from vulnerable ones.
The Bifurcation Nobody Expected
We are not witnessing the end of coding. We are witnessing a bifurcation in the skills companies hire for. While software engineer roles have declined significantly over the past two years, programmer positions have remained relatively stable. The difference matters more than it appears.
If your primary value proposition was writing boilerplate code or fixing straightforward bugs, that work is increasingly automated. Generative AI does those tasks in minutes instead of days. What’s left are tasks AI can’t do reliably yet: transforming vague requirements into specifications, code reviewing AI-generated code for nuanced logic errors, security hardening that goes beyond syntactic validation, connecting components with implicit contextual knowledge.
Day-to-day operations of a mid-level developer have changed profoundly. Where their week in 2022 may have produced maybe 1,000 lines of code implementing a couple of features, now that boilerplate is spit out immediately by AI.
But the real work begins after that: refining specifications, hunting for edge cases the AI missed, conducting what we call “intent auditing” to verify the code does the right thing under real conditions.
Output per engineer has increased, perhaps five or six pull requests weekly instead of two, but each requires significantly more judgment and verification than before.
What Companies Are Actually Hiring For
Data from millions of technical assessments reveals which skills have surged in demand. Programming itself has seen a 54-fold increase in share as an assessed skill. Problem-solving grew 39-fold. Data visualization expanded 35-fold.
These numbers are not random. They reflect a definitive shift from syntax fluency to judgment. Companies are no longer screening primarily for whether candidates know a specific language or framework. They are screening for whether candidates can solve ambiguous problems, visualize outcomes, and make sound technical decisions when multiple approaches exist.
Analytical thinking, problem-solving, and systems thinking have emerged as the new gold standards. These skills are very human: they are outside the realm of what AI can provide.
Sure, an AI model can recommend microservices because that’s been written about a lot in its training data. But AI can’t assess if your team of eight has the Kubernetes knowhow to run a cluster. Or if the operational overhead is worth the slight decrease in latency.
Why Traditional Roles Are Most Vulnerable
Entry-level hiring at top technology firms has declined approximately 60%. The roles most at risk share a common characteristic: they involve work that can be verified objectively and quickly. Maintenance programming, building generic features, and simple bug fixes are examples of this work.
On the other hand, jobs which demand domain knowledge, senior systems design and architecture decisions in unclear environments have stayed steady or increased in importance. This work exists in what I’ll call the “unverifiable space”. There is no objectively right answer and decisions are based on understanding the tradeoffs between business impact, team dynamics, competitive landscape and technical debt at the same time.
A new category of engineer has also emerged: the vibe coder.
These are professionals from adjacent disciplines, marketing managers, product designers, operations specialists, who now use AI to generate functional code for specific tasks like landing pages or internal tools. They may not have degrees in computer science, but AI has decreased the entry level skills required enough that they can write functioning software for limited purposes.
This both democratizes who can create software, and increases the baseline of what professional engineers have to offer vs code creation.
How Hiring Itself Is Changing
The hiring process has historically been constrained by human bandwidth. Interviewers had calendars. Recruiters managed queues. Strong candidates were often lost not because they lacked capability, but because the evaluation process moved too slowly.
The wait times are going away. Automated interviews powered by AI will perform structured, job-level calibrated interviews 24/7. Someone who applies late Sunday night can have their technical screen completed before Monday morning without having to coordinate schedules.
There are some downstream impacts of this which are significant. Applicants are now mass-producing resumes and cover letters using AI, leading to companies receiving an unprecedented number of “high quality” resumes they have difficulty separating. If everyone looks good on paper, being good on paper is no longer useful.
Organizations are pushing towards more validated assessments. Rather than asking for credentials or telling stories, companies are using standardized interviews, work sample tests, and live “think aloud” problem solving to quantify applied knowledge. Interviews can be automated and scored with standardized rubrics to turn qualitative feedback into quantitative data.
Live proctoring is expected. Interview data indicates companies who use proctoring went from 64% in January of 2025 to a high of 77% in July. As of December, this number has settled around 64.5%. Applicants should now expect their interviews to be verified and monitored.
What Workforce-Ready Actually Means Now
Job readiness used to mean preparing for a handful of scheduled interviews. That model is breaking. In an environment where interviews can happen any time and evaluation is continuous, job readiness is no longer a moment you prepare for. It is a state you must maintain.
Candidates are no longer competing as individuals. They are competing as humans working with tools. AI coding assistants and automation platforms are part of standard workflows across engineering roles. The question is not whether someone can complete a task manually, but whether they can produce the correct outcome using available tools and whether they can verify that the outcome is actually correct.
This raises the performance bar in a counterintuitive way. Automation doesn’t ease the candidate’s job. If anything, interviews get harder.
There’s less variance if interviews are structured and data-driven. If you nail the presentation, or wow them in one conversation, that’s probably not enough. You have to consistently perform.
Interviews also become more transparent. There’s less luck involved with a standardized system. More candidates are given a fair shot at proving skill, rather than being eliminated for poor resume formatting or keyword density.
Skills That Are Irreplaceable
Verification skills have become critical. Engineers now spend significantly more time, an estimated 200% increase, on architecture alignment and output validation than they did before AI tools became standard. Reviewing code written by generative AI for hallucinated dependencies, security theater, excessive code reuse and local optimization without considering system-level architecture is another.
Understanding your domain. Generative AI doesn’t know your industry. Whether it’s finance, healthcare, logistics or some other domain, engineers with deep domain expertise have context that AI will not. Institutional knowledge and reasoning about clinical workflows, regulatory limitations, and bottlenecks in a supply chain are examples of what generative AI models will never have.
Practical Career Path Roadmap for Mid-Career Engineers
Stop fighting the tools. Use AI coding assistants for 100% of boilerplate work for two weeks and measure the time savings. Then redirect that time toward systems thinking and design documentation.
Write design docs for everything, regardless of the project size. Document that you considered alternatives, tradeoffs, and consciously made architecture decisions. This shows that you’re an engineer that thinks strategically instead of someone that just implements tasks.
Build proof artifacts, projects where AI handled the majority of implementation, but you provided the critical layer of security validation, integration logic, and edge case handling. These artifacts demonstrate that you are an architect who exercises judgment, not simply someone who types quickly.
AI is Rewriting Software Engineers Roles
AI is not eliminating software engineering jobs. It is rewriting them. The roles being cut are those where value was derived primarily from translating specifications into code. The roles that remain and the new roles emerging require capabilities that sit upstream and downstream of code generation.
Hiring is being rebuilt around demonstrated skill rather than polished narratives. The tools enabling this shift create both opportunity and pressure. Opportunity, because more candidates receive fair evaluation based on actual capability. Pressure, because the bar for what constitutes “capable” has risen significantly.
Advice for engineers that want to make this shift: Learn the basics, practice systems thinking, practice verifying things, and gain domain knowledge.
AI fluency is not about using the tools. It is about knowing when they are right, when they are wrong, and how to provide the judgment that no model can replicate.
The layoffs are real. But they are not the full story. What is being rebuilt, a hiring system grounded in evidence, and an engineering profession redefined around judgment rather than syntax, may ultimately create a more merit-based and durable foundation for the industry.
