Is the era of traditional coding drawing to a close? This pressing inquiry has pervaded discussions among developers this week, following the introduction of formidable new coding models from OpenAI and Anthropic.
In the past week, both OpenAI and Anthropic unveiled their latest coding innovations—GPT-5.3-Codex and Claude Opus 4.6—marking remarkable advancements in AI’s programming prowess.
GPT-5.3-Codex exhibited significantly enhanced performance on coding benchmarks compared to its predecessors, while Opus 4.6 introduced a groundbreaking feature that enables users to deploy autonomous AI agent teams capable of addressing various elements of intricate projects concurrently.
Both models can autonomously write, test, and debug code, often revising their own outputs and fine-tuning features before delivering results to developers.
The launch of these models, especially GPT-5.3-Codex, has ignited a virtual existential debate among software engineers. Central to this discourse is aviral essay by Matt Shumer, CEO of OthersideAI.
Shumer articulated that following these model releases, he perceived a paradigm shift where AI can now autonomously navigate the entire software development lifecycle—generating extensive lines of code, opening applications, testing functionalities, and iterating on outputs—leaving developers to merely articulate their desired outcomes, and step away
He posited that these technological leaps might disturb employment in unprecedented ways, potentially surpassing the disruptions caused by the COVID-19 pandemic.
The essay has elicited a mosaic of reactions. Several tech leaders echoed Schumer’s sentiments, including Reddit co-founder Alexis Ohanian, while others, such as NYU professor Gary Marcus, dismissed it as “weaponized hype.”
Marcus pointed out a lack of empirical data supporting claims that AI could develop complex applications devoid of errors.
In a related commentary, Fortune’s Jeremy Kahnargued that the unique characteristics of coding—such as automated testing—facilitated its potential for complete automation, contrasting with the complexities of fully automating other knowledge-driven professions.
Software Engineers as Pioneers
For numerous engineers, Shumer’s apprehensions resonate with their current experiences. A growing contingent of engineers reports having shifted away from coding in favor of directing AI to produce the necessary code.
While the recent releases signify substantial enhancements, developers assert that the industry has been in the midst of a gradual metamorphosis over the past year, with models acquiring the aptitude to handle increasingly intricate tasks independently.
Though developers at leading technology firms have largely ceased writing code line-by-line, they have not disengaged from software development; rather, they have transitioned to roles as directors overseeing AI systems that perform the actual typing.
This transformation underscores a shift from simply writing code to architecting solutions and shepherding AI tools. Many proponents argue that the new models primarily “burst the bubble” of traditional perceptions regarding AI coding, shedding light on a trend that engineers have been experiencing for months.
During an earnings call this week, Spotify co-CEO Gustav Söderström revealed that the company’s best developers “have not composed a single line of code since December.”
Spotify’s internal architecture employs Claude Code for remote deployment, allowing engineers to instruct AI to rectify bugs or integrate features via Slack during commutes, subsequently merging completed tasks into production prior to their arrival at the office. Söderström highlighted that Spotify introduced over 50 new features in 2025, utilizing these methodologies.
Even within Anthropic, engineers are significantly leveraging their own tools for code development. Boris Cherny, head of Claude Code, disclosed earlier this month that he has not personally written code in over two months. Anthropic previously informed Fortune that between 70% and 90% of the company’s coding is now generated by AI.
The models themselves have ascended to a recursive milestone: they are now substantially contributing to the development of more advanced iterations of themselves.
OpenAI noted that GPT-5.3-Codex “is our first model that played a vital role in creating its successor,” signifying a noteworthy evolution in AI development practices.
Correspondingly, Anthropic’s Cherny mentioned that his team crafted Claude Cowork—a non-technical variant of Claude Code for file management—within a mere week and a half, predominantly utilizing Claude Code itself.
Intriguingly, Cherny revealed that approximately 90% of the code for Claude Code is now authored by Claude Code.
Despite the pronounced productivity enhancements, an array of developers warns that the adoption of these new tools might precipitate burnout. Veteran engineer Steve Yegge asserted that AI tools were overburdening developers through excessive workloads.
In a widely circulated blog post, Yegge recounted instances of abruptly dozing off after extended coding marathons, with colleagues contemplating the installation of nap pods in their workplace.
He contended that the addictive allure of AI coding instruments is incentivizing developers to undertake unsustainable levels of output.
“With a 10x productivity boost, if you equip an engineer with Claude Code, then once they attain fluency, their output will yield value equivalent to nine additional engineers,” he noted. Yet, he cautioned, “constructing solutions with AI necessitates substantial human energy.”
Source link: Inkl.com.
