Demis Hassabis, the CEO of Google DeepMind, is keen to talk about the coding skills of his company’s newest model, Gemini 3.5 Flash. The model has been trained to perform complex agentic coding tasks: translate large code bases from one language to another; find and fix bugs lurking deep in knotty code; and even write entire operating systems from scratch.

Hassabis does not, however, think this spells doom for software developers. “I have no idea why people are going around talking with certainty about that,” Hassabis tells WIRED ahead of the new model reveal at today’s Google’s I/O event.

“Perhaps there is an ulterior motive for putting those messages out; raising money or whatever,” Hassabis says. “From my point of view, from DeepMind and Google's point of view, if engineers are becoming three or four times more productive, then we just [want to] do three or four times more stuff.”

The striking coding abilities of the latest models has led to widespread fear that AI may be on the brink of eliminating programming roles and other white-collar jobs. Executives at some AI companies have predicted widespread job displacement, while some prominent tech companies, including Amazon, Salesforce, and Block, have blamed recent layoffs on the use of AI.

Hassabis thinks that Alphabet, which oversees several companies besides Google, may be well positioned to take advantage of a revolution in software productivity. “I have a million ideas, from lab drug discovery to game design,” he says. “I'd love to have some free engineers to go and do those kinds of things.”

Hassabis says that companies looking to replace developers with AI may be making a big mistake. “I think it's a lack of imagination—and a lack of understanding of what's really going to happen,” he says.

Google revealed a raft of AI stuff at its annual developer event. Through a coding tool called Antigravity, Gemini 3.5 Flash offers frontier coding and reasoning capabilities but is faster and cheaper than the offerings of its competitors, Google says. Gemini 3.5 Pro, a more powerful new version of its flagship model, will debut next month.

The company needs to catch up when it comes to AI coding, which has emerged as a crucial and lucrative application for the latest AI models. Anthropic and OpenAI lead developer adoption with their respective tools, Claude and Codex, according to a 2025 Stack Overflow survey.

The company also demoed an agentic assistant called Spark that lives in Google’s Cloud and has access to its apps. The design is meant to be safer than something like OpenClaw because it has limited access to personal data, Google says.

Other agentic demos included a version of Android with an AI agent built in and a refreshed version of Google Search that uses agentic coding to generate a site or app on the fly in response to a search query.

AI coding has captivated the AI world in recent months, even inspiring hope that models could one day rewrite their own code in a self-improvement loop. Hassabis says it’s possible but doubts that it will immediately lead to superhuman-level AI.

Progress in other areas of science might require AI models to have a deeper understanding of the physical world and even an ability to perform experiments within it, he says.

Even within the seemingly solved world of coding, Hassabis says it is notable that AI has yet to produce a blockbuster app or video game without human help. “I think there's something missing,” he says.