Demis Hassabis, CEO of Google DeepMind, pushed back on using AI as a rationale for developer layoffs while previewing the team’s latest models and tools ahead of Google I/O. He framed recent advances in coding AI as a way to accelerate and broaden engineering work, not to justify headcount reductions — a stance that matters because many firms have cited AI-driven efficiency when trimming staff.
Hassabis showcased Gemini 3.5 Flash as a coding — focused model intended to handle complex, agentic tasks: translating large codebases between languages, locating and fixing deeply hidden bugs, and even writing entire operating systems from scratch. Google also demonstrated Antigravity, a coding tool that combines frontier coding and reasoning capabilities and is presented as faster and cheaper than competitors, and previewed Gemini 3.5 Pro for a public debut next month.
The company emphasized agentic integrations designed to fold AI into developer workflows and cloud tooling rather than standalone consumer apps. Demos included Spark, an assistant operating inside Google Cloud with access to Google apps but restricted access to personal data by design; an Android build with an embedded AI agent; and a version of Search that can generate a site or app on demand, highlighting a push toward embedded, production — oriented tooling.
The timing of these rollouts amplifies a broader industry debate: AI coding is now a primary battleground for developer adoption. A 2025 Stack Overflow survey places Anthropic and OpenAI ahead in developer uptake with Claude and Codex. At the same time, some companies — including Amazon, Salesforce, and Block — have publicly linked recent workforce reductions to efficiency gains from AI, and executive predictions of broad displacement have intensified scrutiny over how productivity gains are deployed.
Hassabis rejected deterministic claims that coding models spell doom for programmers and urged a different approach. If engineers become "three or four times more productive, then we just [want to] do three or four times more stuff," he said, asserting that Alphabet has many ideas for where freed capacity could be applied, from lab drug discovery to game design and other long‑range projects.
He tempered optimism with caution about current limits. Hassabis noted that, despite rapid progress, AI has not yet produced a blockbuster app or video game without significant human involvement and expressed skepticism about an immediate leap to superhuman general intelligence. He argued that further scientific advances may require models with deeper physical — world understanding and the ability to perform experiments, a constraint builders should weigh when setting product road maps.
Sources
Replies (0)
No replies in this topic yet.
