Sundar Pichai dropped a stat today that should have made every software engineer in America sit up. 75% of all new code at Google is now AI-generated. Not helped by AI. Not assisted by AI. Generated by AI and then approved by a human engineer. One year ago that number was 25%. Six months ago it was 50%. Today it is 75%. Pichai announced it in a blog post tied to Google Cloud Next 2026 in Vegas and the number lands harder than any benchmark or product launch has all year. Because this is not a prediction about the future of software engineering. This is the present tense. At the most valuable technology company in the world, AI is now the primary author of new code and humans are the reviewers. The question is not whether this trend spreads. The question is how fast.

The trajectory is what makes this story hit different. In April 2025 Sundar Pichai said AI generated "well over 30%" of Google's code. By late 2025 Google's CFO put the number at "nearly half." This fall Pichai updated it to 50%. Today the number is 75%. That is a tripling in twelve months. And the rate of acceleration is increasing, not slowing down.

Google is not alone. This is the part of the story that makes it impossible to write off as a one-company outlier.

Anthropic is at nearly 100%. The company building Claude now writes essentially all of its code with AI assistance. This tracks. Anthropic engineers reportedly use Claude for 60% of their work today, up from 28% a year ago, with roughly 50% productivity gains. The company ships 60 to 100 internal releases per day. When the tool becomes good enough to build itself, the entire company pace changes.

Microsoft was at 30% a year ago. Satya Nadella disclosed in April 2025 that roughly 30% of Microsoft's code was being written by AI. That number is almost certainly higher now, likely closer to 50% based on their shipping pace.

Snap is at 65%. Just last week Snap laid off 1,000 employees citing "rapid advancements in artificial intelligence" that allow smaller teams to achieve the same output. 65% of Snap's new code is generated by AI. Stock jumped 11%.

Meta is targeting 75% of engineers. An internal document reviewed by Business Insider shows Meta expects 65% of engineers in its creation org to write more than 75% of their committed code using AI in the first half of 2026. The goal is not ambitious by 2025 standards. The goal is catch-up.

Google revealed something else today that makes the percentage even more meaningful. Pichai said a recent complex code migration was completed 6x faster than was possible a year ago because agents and engineers worked together. The speed is what matters. Not just that AI writes the code, but that AI finishes projects in days that used to take months.

Because the people writing the code are becoming optional.

For twenty years the bottleneck at every software company was the same. You could not scale engineering output without hiring more engineers. Every feature required a person to write it. Every bug required a person to fix it. Every migration required a team to execute it. The size of your engineering team determined the speed at which your company could move. That constraint has defined how Silicon Valley builds companies since the dot com era.

That constraint is now broken.

Google just completed a complex code migration 6x faster than a year ago. Not with more engineers. With the same or fewer engineers using AI agents. Snap just laid off 25% of their planned headcount because AI can do the work. Stanford's 2026 AI Index report released last week showed employment for US software developers aged 22 to 25 has dropped nearly 20% since 2022. The decline is not evenly distributed. Senior engineers are still in demand. Junior engineers, the ones who used to learn by doing the tasks AI now does, are the ones getting squeezed out.

This is not a future problem. This is happening right now in every major engineering org in the industry. Google at 75%. Snap at 65%. Anthropic at near 100%. If your job is to write code for a living, the number of companies that still need a human to do that as the primary activity is shrinking by the quarter.

⚡ The Vibe Check: A year ago the question was whether AI could really write production code that ships. Today the question is what engineers do when AI is writing most of it. Pichai framed the 75% stat as a productivity milestone. But when the CEO of one of the largest engineering organizations on earth says AI writes three quarters of the new code, that is not a productivity update. That is a quiet announcement about what software engineering is becoming.

If you write code for a living, the advice has not changed but the urgency has. Learn to use AI tools daily. Not occasionally. Not when it feels convenient. Daily. The engineers surviving this shift are the ones directing AI agents, reviewing their output, and architecting the systems the AI then fills in. That is the new skill set. The engineers who refused to adopt AI tools in 2024 are the ones being displaced in 2026.

If you manage engineers, your hiring model is about to change. The 2019 playbook was hire more bodies to go faster. The 2026 playbook is hire fewer bodies with better AI workflows. Companies that figure this out first will ship faster at lower cost than their competitors. The laggards will look expensive and slow by comparison.

If you are watching the industry, pay attention to what happens next at Google. 75% today is 85% by Q4. 90% by next year. The companies that hit 100% first will define what software development looks like for the next decade. And the companies that hit 100% earliest are the AI labs themselves. Anthropic is there. OpenAI is probably close. Google is accelerating toward it.

The code is being written. The humans are becoming curators.

Sundar Pichai confirmed today that 75% of all new code at Google is now AI-generated, up from 25% a year ago. Anthropic is at nearly 100%. Snap is at 65%. Meta is targeting 75% of engineers by mid-2026. Microsoft is ramping hard. The stat lands in the middle of the week that Snap laid off 1,000 employees citing AI and Stanford released data showing young developer employment down 20% since 2022. The software engineering profession is not ending. It is transforming in real time. The engineers who adapt become orchestrators of AI agents. The engineers who do not become the next category of worker replaced by their own tools. Three quarters of Google's code is written by AI today. What is the number at your company? And what is it going to be in six months?

Stay building. 🤖