This article explores how software engineers are responding to the disruption caused by AI. Many, like Matt, are intentionally avoiding AI tools during personal projects to maintain their core coding skills, as their jobs shift from coding to reviewing AI-generated code.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
Since ChatGPT's launch, over 600,000 tech jobs in the US have disappeared, and computer science graduate unemployment has risen to 7%. Software engineers face career reinvention: sticking to manual coding, shifting to AI evaluation, or leaving the industry. Experts agree that coding skills are depreciating, while the ability to evaluate AI code, define problems, and design systems is becoming the new core of value.
Since the release of OpenAI's ChatGPT in 2022, the software engineering profession has faced unprecedented disruption. Google reports that 75% of its code is now written by AI, and engineer Matt finds his role shifting from coding to reviewing AI-generated code, stating bluntly, 'I'm trying not to use AI.' Dr. Bouke Klein Teeselink, assistant professor at King's College London, notes: 'The skill of writing code is becoming obsolete, while the ability to evaluate AI code, spot bugs, and understand errors is becoming crucial.'
The job market has changed drastically. According to Layoff.fyi, over 600,000 US tech workers have lost their jobs. New York Fed data shows that the unemployment rate for computer science graduates in 2024 rose to 7% (from 6.1% the previous year), with underemployment exceeding 19%. Indeed tech job postings in the US have fallen 36% compared to 2020. Despite fewer positions, experts emphasize that software engineers are not obsolete: Wharton professor Ethan Mollick argues that value now lies in defining problems, designing systems, and effectively directing AI tools.
Engineers are adopting different strategies. Matt avoids using AI in personal projects, hand-coding every line to keep his skills sharp. He fears over-reliance on AI could weaken his abilities, but his current boss demands more AI use, making him feel 'dark' about the future. In contrast, after being laid off from Mailchimp in 2024, George Dover turned to learning AI-generated code and carefully evaluating its errors, redundancies, and bugs; after 400 applications, he landed an AI-focused software engineering position.
Others consider collective action or leaving entirely. Dover briefly became a substitute kindergarten teacher, thinking 'What else fits me?' But the transition isn't easy: Dover found that AI code sometimes requires more debugging time, 'AI can lead you down a rabbit hole.' Klein Teeselink points out that while non-technical workers can write code, verifying AI code still requires professional skills, which preserves demand for engineers.
More than a decade ago, programming was seen as the core of economic opportunity. In 2013, Obama launched a $4 billion 'Computer Science for All' initiative, with Zuckerberg and Gates also promoting it, and coding bootcamps boomed. Now these efforts are disrupted by AI—the automation of code writing has far exceeded expectations.
The future remains uncertain. Harvard professor David Malan notes that AI operational costs are extremely high (OpenAI spent $8 billion last year, Anthropic about $3 billion), and these costs will eventually be passed to customers, so companies are unlikely to rely entirely on AI but will seek a 'AI-augmented software engineer' balance. Brown professor Shriram Krishnamurthi adds that AI only started generating high-quality code last year, and the need for code review will filter out prepared engineers, but 'the career landscape in two years is hard to predict.'
This article is based on an in-depth report from The Guardian, incorporating statements from multiple engineers and academics, as well as public statistics (Layoff.fyi, New York Fed, U.S. Bureau of Labor Statistics, Reuters). Event details come from interviewees' own accounts, and data is traceable, but future projections are expert inferences and involve uncertainty.
Software engineers' careers are being reshaped by AI, with core skills shifting from writing code to evaluating AI output, system design, and problem definition. Adapters need to proactively learn the limitations and advantages of AI tools, while clinging to old skills or completely exiting both carry risks. The high cost of AI may delay full replacement, but the career structure has undergone irreversible change.
Primary report
Primary source