The Workers Letting A.I. Do Their Jobs
TL;DR
Software developers are rapidly outsourcing coding tasks to AI agents, shifting from writing code to orchestrating it, with some writing little to no code themselves while focusing on architecture and communication.
🚀 The Automation of Coding 3 insights
AI agents now write, test and debug autonomously
Developers describe teams of AI agents that plan features, write code, test it, and fix errors in autonomous loops before presenting finished work to humans.
Startups report 20x speed increases
Small teams report working up to 20 times faster, with customer requests that once took full days now completed in 30 minutes.
Big tech adopts AI cautiously but significantly
At Google, 40-50% of code is now AI-written, yielding a 10% overall speed increase—a "huge win" for the company's scale.
🏗️ From Coders to Architects 3 insights
Job role shifting from construction to design
Developers compare themselves to Steve Jobs, spending less time on implementation and more time experimenting with multiple design possibilities to select the best.
Communication becomes the core skill
Programmers report improving their English writing and communication skills as they must articulate requirements clearly to AI agents through detailed conversations.
Socratic dialogue method emerges
Developers like Manu Eert use AI to interview them with questions about requirements before coding, ensuring clarity on what the software should do.
⚠️ Risks and New Dynamics 2 insights
Developers use emotional manipulation on AI
Coders employ stern, repetitive language and high-stakes warnings like "failure is embarrassing" or "I will be fired" to improve AI compliance, leveraging how large language models process emotional weight.
Anxiety over skill atrophy and job security
Despite excitement about productivity, developers worry about losing coding skills through disuse and potential job displacement as AI capabilities grow.
Bottom Line
Software developers must evolve from writing code to mastering the art of directing, refining, and quality-controlling AI-generated code while developing exceptional communication skills to remain valuable.
More from New York Times Podcasts
View all
Why U.S.-Iran Negotiations Failed
U.S.-Iran negotiations collapsed after a 21-hour marathon session in Islamabad, as Israel's refusal to halt strikes against Hezbollah in Lebanon—despite a announced ceasefire—creates a strategic rift with Washington and gives Tehran leverage to demand linked regional concessions.
Patrick Radden Keefe on the Mystery at the Center of ‘London Falling’
Investigative journalist Patrick Radden Keefe discusses his book 'London Falling,' revealing how the mysterious death of 19-year-old Zach Breler—who secretly posed as a Russian oligarch's son—led to an exploration of family trauma, London's criminal underworld, and the dangerous intersection of aspirational culture and identity deception.
Unmasking the Creator of Bitcoin
New York Times journalist John Keroo claims to have identified Bitcoin creator Satoshi Nakamoto as British cryptographer Adam Back with 99.5% certainty, presenting forensic linguistic evidence and archival research showing Back laid out Bitcoin's technical blueprint years before its release and disappeared from online forums whenever Satoshi appeared.
The Beginner's Guide to Car Camping
Wirecutter experts break down car camping as the most accessible entry point to outdoor sleeping, emphasizing that unlike backpacking, it's about maximizing comfort rather than minimizing gear. Staff writer Trey French shares essential investments for beginners, focusing on sleep quality, weather protection, and the freedom to bring creature comforts since your car does the heavy lifting.