Is conventional coding useless? That’s the query many builders have been asking themselves this week following the launch of highly effective new coding fashions from OpenAI and Anthropic.
Final week, OpenAI and Anthropic dropped their respective coding fashions—GPT-5.3-Codex and Claude Opus 4.6—each of which represented important leaps in AI coding capabilities. GPT-5.3-Codex confirmed markedly increased efficiency on coding benchmarks than earlier fashions, whereas Opus 4.6 launched a function that lets customers deploy autonomous AI agent groups that may sort out totally different features of complicated tasks concurrently. Each fashions can write, check, and debug code with minimal human intervention—even iterating on their very own work and refining options earlier than presenting outcomes to builders.
The releases—particularly GPT-5.3-Codex—sparked one thing of a web-based existential disaster amongst software program engineers. On the coronary heart of it was a viral essay written by Matt Shumer, CEO of OthersideAI. Shumer mentioned that “something clicked” following the mannequin releases and described AI fashions now dealing with all the growth cycle autonomously—writing tens of 1000’s of strains of code, opening purposes, testing options, and iterating till glad, with builders merely describing desired outcomes and strolling away. He proposed that the advances meant that AI may disrupt jobs extra severely than the COVID-19 pandemic.
The essay drew combined reactions. Some tech leaders agreed, together with Reddit co-founder Alexis Ohanian, however others, together with NYU professor Gary Marcus, criticized it as “weaponized hype.” (Marcus famous that Shumer supplied no knowledge supporting claims that AI can write complicated apps with out errors.) Fortune’s Jeremy Kahn additionally argued that it was coding’s distinctive traits—like automated testing—that made it simpler to completely automate, whereas the automation of different knowledge-work fields could also be extra elusive.
Software program engineers as early adopters
For a lot of engineers, a few of Shumer’s warnings simply replicate their present actuality. Many engineers say they’ve stopped coding fully, as a substitute counting on AI to put in writing code at their course.
Whereas the brand new releases do symbolize significant enhancements, builders additionally mentioned the trade has been present process a gradual transformation over the previous yr as fashions turned succesful sufficient to deal with more and more complicated duties autonomously. Whereas builders at main tech firms have largely stopped writing code line-by-line, they haven’t stopped constructing software program—they’ve change into administrators of AI programs that do the typing for them. The ability has reworked from writing code to architecting options and guiding AI instruments. The brand new fashions, some argue, primarily “burst the bubble” round AI coding by making folks exterior coding conscious of a pattern engineers have been experiencing for months.
Throughout its earnings name this week, Spotify co-CEO Gustav Söderström mentioned the corporate’s greatest builders “have not written a single line of code since December.” The streaming big’s inside system makes use of Claude Code for distant deployment, permitting engineers to instruct AI to repair bugs or add options by way of Slack on their telephones throughout their commute, then merge accomplished work to manufacturing earlier than reaching the workplace. Söderström mentioned Spotify shipped over 50 new options in 2025 utilizing these workflows.
Even inside Anthropic, engineers are closely counting on their very own instruments to put in writing new code. Boris Cherny, head of Claude Code, mentioned earlier this month that he hasn’t written code in over two months. Anthropic beforehand instructed Fortune that 70% to 90% of the corporate’s code is now AI-generated.
The fashions themselves have additionally reached a recursive milestone: They’re now materially serving to to construct extra superior iterations of themselves. OpenAI mentioned GPT-5.3-Codex “is our first model that was instrumental in creating itself,” a big shift in how AI growth works. Equally, Anthropic’s Cherny mentioned his workforce constructed Claude Cowork—a non-technical model of Claude Code for file administration—in roughly per week and a half, largely utilizing Claude Code itself. Even for Claude Code, Cherny mentioned about 90% of its personal code is now written by Claude Code.
Regardless of the productiveness positive aspects, some builders are additionally warning that the brand new instruments may end in burnout. Steve Yegge, a veteran engineer, mentioned that AI instruments had been draining builders by way of overwork.
In a extensively shared blogpost, Yegge described falling asleep all of the sudden after lengthy coding classes and colleagues contemplating putting in nap pods at their workplace. The addictive nature of AI coding instruments, he argues, is pushing builders to tackle unsustainable workloads. “With a 10x boost, if you give an engineer Claude Code, then once they’re fluent, their work stream will produce nine additional engineers’ worth of value,” he wrote. However “building things with AI takes a lot of human energy.”
