Phoenix 24
The engineer is no longer alone at the keyboard.
Mountain View, April 2026.
Sundar Pichai has confirmed that artificial intelligence now generates 75% of Google’s new code, a figure that marks one of the clearest signs yet that software development has entered a structural transformation. The company’s engineers still review and approve the output, but the center of gravity has shifted. Coding is no longer defined only by human authorship. It is becoming a process of supervision, validation and orchestration between engineers and intelligent systems.
The number matters because of its speed. Google had already reported that AI was generating around half of its new code months earlier, and the jump to 75% shows that adoption is accelerating inside one of the most technically advanced companies in the world. This is not a laboratory experiment or a marginal productivity tool. It is becoming part of the operating logic of Big Tech.
The change also redefines the role of the software engineer. The traditional image of a developer writing every line manually is giving way to a new model where engineers guide AI agents, review generated solutions, detect errors and decide what enters production. In that environment, technical authority does not disappear, but it migrates. The most valuable skill becomes not only writing code, but knowing how to judge code produced at machine speed.
Google’s internal use of AI also functions as a strategic signal to the enterprise market. If the company can demonstrate that its own infrastructure, Gemini models and agentic workflows accelerate software production, it strengthens the commercial case for selling those tools to businesses. Google is not simply using AI to code. It is turning its internal transformation into a product argument.
The productivity promise is enormous, but so are the risks. AI-generated code can accelerate migrations, reduce repetitive labor and expand engineering capacity, but it can also introduce hidden vulnerabilities, dependency errors, security flaws or architectural inconsistencies if review processes weaken. The more code machines produce, the more important human governance becomes. Automation does not eliminate responsibility; it concentrates it at the point of approval.
This shift also carries labor implications. Engineers are not being replaced in a simple or immediate sense, but their work is being reorganized. Tasks once considered central may become automated, while judgment, system design, debugging and risk evaluation become more critical. The profession is moving from execution toward oversight, and that transition will create winners, anxiety and new hierarchies inside technical teams.
For the broader technology industry, Google’s figure becomes a benchmark. If one of the world’s most sophisticated engineering organizations can operate with most new code AI-generated, competitors will face pressure to match that productivity curve. Microsoft, Meta, Amazon and startups across the ecosystem will not treat this as a curiosity. They will read it as a competitive threshold.
The deeper question is quality. More code does not automatically mean better software. In complex systems, speed can create technical debt if architectural discipline fails to keep pace. The challenge for Google is not proving that AI can generate code. That point is already settled. The real test is whether AI-generated code can remain secure, maintainable and aligned with long-term product integrity at massive scale.
What emerges is a new software economy where authorship becomes distributed. Human engineers define intent, machines generate implementation and organizations decide what level of autonomy they are willing to trust. This is not the end of programming. It is the beginning of a different programming culture, one where the keyboard becomes less important than the review layer around it.
Pichai’s disclosure does not simply describe an internal metric. It reveals the direction of the industry. The future of software will not be built only by humans or only by machines, but by systems where both operate in continuous negotiation. The companies that master that relationship will define the next decade of digital infrastructure.
Code is no longer only written, it is governed.
El código ya no solo se escribe, también se gobierna.