AI coding tools from OpenAI and Anthropic ignite debate over the future of programming jobs.
The release of two advanced AI coding models has reignited a major conversation in the global tech industry: is traditional programming slowly becoming obsolete?
On February 5, OpenAI unveiled GPT-5.3-Codex, while Anthropic introduced Claude Opus 4.6. The near-simultaneous launches have intensified arguments about how far artificial intelligence can go in replacing human developers.
The discussion gained momentum after Spotify disclosed that its most senior engineers have not written code manually since December.
During the company’s February 10 earnings call, co-CEO Gustav Söderström explained that developers now use an internal AI system called Honk, powered by Claude Code.
He said engineers can instruct the AI through Slack even while commuting to fix bugs or add features to the iOS app. Once completed, a new version of the app is delivered directly to their phones for review and production approval.
The claim has fueled speculation that AI coding assistants are rapidly transforming workplace practices.
Entrepreneur Matt Shumer added to the debate with a viral essay titled “Something Big Is Happening,” viewed more than 80 million times on X.
In the piece, he argued that AI can now handle all his technical work and predicted disruption “much bigger than Covid.”
NYU emeritus professor Gary Marcus criticised the essay, describing it as exaggerated and warning that AI systems still struggle with reliability and security flaws.
OpenAI says GPT-5.3-Codex played a role in its own development, debugging training runs and identifying issues during testing.
Anthropic claims Claude Opus 4.6 now ranks first on financial analyst benchmarks and delivers stronger performance across large codebases.
These advances suggest AI coding systems are becoming more capable, especially in complex technical tasks.
Despite the optimism, recent research highlights growing risks.
A study published on Arxiv found that 88 percent of developers reported increased technical debt when using AI tools. More than half said the generated code often appeared correct but proved unreliable.
Separate findings linked AI adoption to higher workplace pressure and burnout.
Software engineer Steve Yegge warned that engineers relying heavily on rapid AI-driven coding may only sustain peak productivity for about three hours daily before performance drops.
GitClear’s analysis of 153 million lines of code revealed code duplication rose by 48 percent between 2021 and 2024, while refactoring declined sharply in AI-assisted projects.
Some tech leaders believe major change is coming fast.
Microsoft CEO Mustafa Suleyman predicted that most white-collar tasks could be automated within 18 months. Elon Musk has suggested that coding as a profession may effectively end by 2026.
Marcus remains cautious. He believes AI will eventually reshape labour markets, but not within the next one or two years.
_1769956161.jpg)
0 Comments