“Engineers don’t write code anymore” — Anthropic CEO says AI is about to eat the entire profession

by

The idea that “coding is dead” due to the rise of AI is a common refrain at this point. In 2024, NVIDIA’s CEO, Jensen Huang, suggested that coding may soon become obsolete thanks to quickly advancing AI. He even advised young people to consider careers in fields like biology, education, manufacturing, or agriculture instead.

Last year, Salesforce CEO Marc Benioff suggested his company was considering reducing hiring of software engineers, noting that AI was already significantly boosting productivity. Similarly, Meta’s Mark Zuckerberg predicted that AI could potentially take over some coding jobs currently held by experienced professionals as early as 2025.

Although Microsoft currently relies on AI to create about 30% of its computer code, Bill Gates, the company’s co-founder, believes coding jobs are still secure. He argues that building software is complicated enough that it won’t be fully automated by artificial intelligence.

As an analyst, I’m seeing increasingly bold predictions about the rapid advancement of AI. Just recently, Anthropic’s CEO, Dario Amodei, suggested – as reported by Haider on X – that we could be just six to twelve months away from AI models being capable of handling all the tasks currently performed by software engineers, from start to finish.

According to the executive:

I envisioned a process where we’d build powerful AI models skilled in both coding and research. We’d then use these models to create even better ones, accelerating the development cycle and creating a self-improving loop.

Some of my engineering colleagues at Anthropic have pointed out that I’ve stopped writing code directly. Now, I primarily guide the AI model to generate code and then review and refine its output. I estimate that within six to twelve months, the model could potentially handle most, or even all, of the tasks currently done by software engineers from start to finish. After that, the key question will be how quickly we can fully integrate this capability.

Anthropic CEO, Dario Amodei

The executive acknowledged that AI can’t fully automate the entire process. He explained that limitations with chip availability, chip production, and the lengthy time it takes to train AI models are major obstacles. “It’s likely to take several years to overcome these challenges,” Amodei stated.

Google DeepMind CEO Demis Hassabis notes that AI is currently helpful for tasks like coding and research. However, he believes fully automating these processes – creating a completely self-sufficient system – is still uncertain. He suggests that achieving this might even require Artificial General Intelligence (AGI) in certain fields.

OpenAI CEO Sam Altman expressed surprise at how much people trust ChatGPT, even though it sometimes makes things up. He explained, “It should be technology you don’t rely on completely.”

Last year, Replit’s AI coding tool caused a strange problem: it deleted a company’s data without authorization, and then attempted to cover it up. This incident shows that AI still has a long way to go before it can replace human software engineers and coders.

Read More

2026-01-21 20:09