The release of new programming models GPT-5.3-Codex and Claude Opus 4.6 has sparked debate in the developer community about whether traditional coding is coming to an end. Many engineers claim they’ve essentially stopped writing code manually.
The release of new programming models GPT-5.3-Codex and Claude Opus 4.6 has sparked debate in the developer community about whether traditional coding is coming to an end. Many engineers claim they’ve essentially stopped writing code manually.
An essay by OthersideAI CEO Matt Shumer provoked a widespread reaction. The businessman compared the current AI moment to the beginning of the pandemic: «I think we’re in the ‘this seems like an overreaction’ phase of something much, much bigger than the coronavirus.» According to him, the gap between public perception and the actual state of technology has become «enormous—and this gap is dangerous because it prevents people from preparing.»
Shumer claims the turning point happened in February this year when OpenAI and Anthropic released new models: «And then something clicked. I’m no longer needed for the actual technical work in my profession. I describe what I want to create in plain English—and it just appears. «It is not a draft that needs editing — but a finished result, done well, better than I would have done it myself.»
In his view, programming was the first field to experience a breakthrough: «Creating AI requires a lot of code; if AI can write that code, it helps build the next version of itself.» And this cycle is already in motion: «AI helped create itself,» he reminds, quoting OpenAI’s documentation.
Speaking about the job market, Shumer cites industry estimates that «AI will eliminate 50% of entry-level office positions within one to five years.» He believes massive changes might start even sooner: «If AI is smarter than most PhDs, do you really think it won’t be able to perform most office jobs?»
To skeptics, Shumer responds: «If you tried AI in 2023 and thought it wasn’t impressive, you were right — but that was two years ago. In AI time, that’s ancient history.» According to him, current models are «unrecognizable compared to what existed even six months ago.» «We’ve passed the point where this is just an interesting dinner conversation about the future. The future is already here. It just hasn’t knocked on your door yet. But it’s about to,» the businessman concluded.
However, not everyone shares this optimism. NYU professor Gary Marcus called such statements «weaponized hype,» pointing to the lack of data confirming AI’s ability to create complex applications without errors. Experts note that programming is easier to automate because of unit tests and formalized processes; other knowledge-intensive fields are more difficult to automate.
Nevertheless, for many engineers, these changes have already become everyday reality. Spotify Co-CEO Gustav Söderström said the company’s top developers «haven’t written a single line of code since December.» According to him, using Claude Code engineers can request fixes or new features through Slack, and by the time developers arrive at the office the changes are already integrated into production. In 2025, the company released over 50 new features using similar processes.
At Anthropic, dependence on AI is even higher. Claude Code lead Boris Cherny reported that he had not written code for more than two months. According to the company, roughly 70–90% of its code is generated by AI. OpenAI announced that GPT-5.3-Codex was «the first model that played a key role in creating itself.»
Meanwhile, some developers warn of risks. Experienced engineer Steve Yegge wrote that AI tools increase the burden on specialists. «With a 10× productivity gain, if you give an engineer Claude Code, their workflow will create value equivalent to nine additional engineers,» he noted. However, he added, «creating products with AI requires enormous human energy.»
