Coding After Coders: AI Agents Are Replacing Traditional Programming
Key Facts
- What: A major New York Times Magazine investigation reveals that programmers at Google, Amazon, Microsoft and Apple now spend most of their time directing AI agents in plain English instead of writing code themselves.
- Who: The piece is based on interviews with more than 70 software developers across leading tech companies plus prominent voices including Anil Dash, Thomas Ptacek, Steve Yegge and Simon Willison.
- When: Published March 12, 2026, capturing the current state of AI-assisted development in 2026.
- How: Developers describe requirements to AI agents, review the agents’ plans, then let the agents generate, test and iterate on code autonomously.
- Unique Advantage: Unlike law or other professions, coding allows AI outputs to be automatically tested against reality, giving programmers a significant edge in adopting the technology.
Lead paragraph
Programmers at the world’s biggest tech companies have largely stopped writing code. Instead, they now spend their days talking to AI agents in plain English, reviewing the agents’ plans, and turning them loose to generate and test software — a profound shift that marks the end of computer programming as it has been practiced for decades, according to a sweeping New York Times Magazine investigation published today.
The article, titled “Coding After Coders: The End of Computer Programming as We Know It,” paints a picture of Silicon Valley that feels “deeply, deeply weird” to traditional developers. Rather than crafting algorithms line by line, engineers now act more like orchestra conductors, directing fleets of AI systems that do the actual implementation work.
The New Workflow
Clive Thompson’s deeply reported piece details a fundamental change in daily developer life. Engineers describe what they want in natural language. The AI produces a plan. The engineer reviews and refines that plan. Then the agent is released to write code, run tests, fix failures, and iterate until the software meets requirements.
This represents a dramatic departure from even the recent past when developers used AI as a sophisticated autocomplete tool. Today’s reality is closer to full delegation.
Simon Willison, a prominent independent developer and blogger, told Thompson that programmers may actually have it easier than many other professionals facing AI disruption. “Given A.I.’s penchant to hallucinate, it might seem reckless to let agents push code out into the real world,” Willison said. “But software developers point out that coding has a unique quality: They can tether their A.I.s to reality, because they can demand the agents test the code to see if it runs correctly. ‘I feel like programmers have it easy,’ says Simon Willison… ‘If you’re a lawyer, you’re screwed, right?’ There’s no way to automatically check a legal brief written by A.I. for hallucinations — other than face total humiliation in court.”
Optimism Amid Disruption
The overwhelming tone from the 70+ developers interviewed was optimistic. Many believe the Jevons paradox — where increased efficiency drives greater overall demand — could actually expand the need for programming talent rather than shrink it. As software becomes dramatically cheaper and faster to create, organizations are expected to build far more applications than previously imagined.
This mirrors historical shifts in programming. As one observer noted in related commentary, the first programmers physically wired circuits. They were replaced by those who toggled binary switches, who were then replaced by assembly programmers, and so on. Each generation saw the previous skill set become obsolete while creating more valuable work at a higher level of abstraction.
However, not every voice is enthusiastic. One Apple engineer, speaking anonymously for fear of repercussions, lamented the loss of craft: “I believe that it can be fun and fulfilling and engaging, and having the computer do it for you strips you of that.” The request for anonymity highlights how corporate culture at major AI-adopting companies may be suppressing dissenting opinions about the changes.
Competitive Landscape
Major technology companies are all embracing this new paradigm at scale. Google, Amazon, Microsoft and Apple have integrated advanced AI coding agents into their development workflows. The tools allow engineers to operate at a much higher level of abstraction, focusing on system design, requirements and architecture rather than implementation details.
This shift comes as AI coding tools have reached a level of maturity where they can handle complete features, debug complex issues, and even write tests autonomously. The New York Times piece arrives at a moment when many in the industry are still processing what these changes mean for job roles, required skills and career trajectories.
Impact
For developers, this changes everything. Traditional coding skills are being supplemented — and in many cases replaced — by new abilities: prompt engineering at scale, plan evaluation, system orchestration, and AI output validation. The most valuable engineers may soon be those who best understand how to direct and constrain powerful AI systems rather than those who write the cleanest code by hand.
“I feel like programmers have it easy. If you’re a lawyer, you’re screwed, right?”
This quote from Simon Willison, highlighted in the New York Times piece, captures a crucial distinction: software engineering has built-in verification mechanisms that many other knowledge professions lack. That reality may protect programming jobs from the most extreme disruption scenarios while fundamentally transforming what those jobs actually entail.
The changes also carry emotional weight for many in the field. For developers who entered the profession because they loved the tactile, creative act of writing elegant code, the shift to AI orchestration can feel like losing a core part of their identity and daily satisfaction.
What’s Next
The industry appears to be in the early stages of inventing an entirely new programming paradigm built around AI systems. As one developer observed, the transition resembles the leap from desktop to internet-era development, where the entire stack changed even though the fundamental layers remained.
Looking ahead, experts anticipate continued rapid evolution of AI coding agents. Capabilities that seem advanced today will likely become table stakes within 12-18 months. Companies that most effectively integrate these tools into their workflows may gain significant advantages in speed to market and engineering efficiency.
The piece also raises important questions about the future shape of software teams. Will smaller teams be able to accomplish what previously required dozens of engineers? How will computer science education need to adapt? What new skills will define the next generation of elite developers?
While the article stops short of declaring the end of programming jobs, it makes clear that programming as a daily practice has already been transformed — perhaps permanently.
Emotional Stakes
For millions of software developers worldwide, this moment represents both opportunity and disruption. Those who adapt quickly to the new reality of AI-directed development may find their productivity and impact multiplied. Those who resist the change risk being left behind as the industry rewrites its own rules in real time.
The New York Times investigation arrives at a pivotal moment when the AI industry is moving faster than ever. What seemed like speculative futurism just two years ago has become the daily reality inside the world’s most important technology companies.