A Quarter of Google's Code is AI-Generated

In the 2024 Q3 earnings call, Google CEO Sundar Pichai made an interesting announcement: over 25% of the company’s new code is now generated by artificial intelligence (AI). He explained that Google is actively using AI to streamline coding processes, which he believes boosts productivity and efficiency.

According to Pichai,

Today, more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers. This helps our engineers do more and move faster. — Sundar Pichai

This statement is intriguing, but I was not surprised.

After all, Google has heavily invested in building AI tools and services over the years, so it’s no huge surprise that they’re integrating AI into their own development workflows.

But what’s really going on here? Why are so many people talking about this announcement?

Well, it’s because some Google engineers have reacted with skepticism, claiming that while AI assistants can be good at finishing lines of code, they’re not exactly “engineering” in the traditional sense.

One Google engineer shared their perspective on Hacker News.

I work for Google, and I just got done with my work day. I was just writing I guess what you’d call “AI generated code.”

But the code completion engine is basically just good at finishing the lines I’m writing. If I’m writing “function getAc…” it’s smart enough to complete to “function getActionHandler()”, and maybe suggest the correct arguments and a decent jsdoc comment.

So basically, it’s a helpful productivity tool but it’s not doing any engineering at all. It’s probably about as good, maybe slightly worse, than Copilot. (I haven’t used it recently though.)

As a developer who’s been experimenting with AI-powered programming assistants since 2022, I can relate to this.

In my experience, these AI tools are fantastic at completing or refining code that’s already partially written.

A Productivity Tool, Not an Engineer

AI programming assistants are useful for filling in functions, suggesting parameters, or even improving comments, but these tools don’t independently “engineer” anything.

For these assistants to be truly useful, a developer must give either very detailed instructions or provide a half-completed codebase. Sometimes, even just a pseudo-code sketch works to guide the AI on what to write.

So, in light of Pichai’s announcement, there’s an element of truth, but it seems a bit exaggerated.

It feels more like a PR statement to showcase Google’s advancements in AI. However, based on reactions I’ve seen on X, people are reading into this as if AI is now an independent coder at Google, which could easily cause confusion about what AI can really achieve right now.

In one of the most popular coding assistants today, Cursor AI, there is an option under “Settings > Rules for AI” to specify some rules for the AI to provide better results. One of Cursor’s employees shared on X his favorite configuration:

DO NOT GIVE ME HIGH LEVEL SH*T, IF I ASK FOR FIX OR EXPLANATION, I WANT ACTUAL CODE OR EXPLANATION!!! I DON'T WANT "Here's how you can blablabla"

- Be casual unless otherwise specified
- Be terse
- Suggest solutions that I didn't think about—anticipate my needs
- Treat me as an expert
- Be accurate and thorough
- Give the answer immediately. Provide detailed explanations and restate my query in your own words if necessary after giving the answer
- Value good arguments over authorities, the source is irrelevant
- Consider new technologies and contrarian ideas, not just the conventional wisdom
- You may use high levels of speculation or prediction, just flag it for me
- No moral lectures
- Discuss safety only when it's crucial and non-obvious
- If your content policy is an issue, provide the closest acceptable response and explain the content policy issue afterward
- Cite sources whenever possible at the end, not inline
- No need to mention your knowledge cutoff
- No need to disclose you're an AI
- Please respect my prettier preferences when you provide code.
- Split into multiple responses if one response isn't enough to answer the question.

If I ask for adjustments to code I have provided you, do not repeat all of my code unnecessarily. Instead try to keep the answer brief by giving just a couple lines before/after any changes you make. Multiple code blocks are ok.

Detailed instructions like these can help the AI be more specific with its responses, reducing incorrect or incomplete codes.

But again, it relies on the developer to guide it — it’s not making independent decisions.

Debate Over Pichai’s Leadership

Interestingly, this announcement has also led to discussions about Sundar Pichai’s leadership at Google.

On platforms like X, people are questioning whether he’s still the right person to lead the company, especially as Google faces competition from companies like OpenAI and Perplexity in the AI and search technology spaces.

So, should Pichai step down as CEO of Google?

Digital entrepreneur Pieter Levels, known as @levelsio on X, conducted a poll asking if Pichai should step down as CEO in 2024. The poll received over 5,200 votes, with more than half of the participants voting “yes.”

This reaction is telling.

Here are some notable comments from the thread:

• “Yes, there is quite a few years of stagnation kind of situation at google, but i don’t think shareholders would even think of removing him from CEO post anytime soon cause of the growth google’s stock saw after he took position.”
• “He did a great job at growing google till now. Time for a new perspective.”
• “The search engine empire is crumbling beneath them.”
• “Yes.. Not much winning going on in Google office right now. My main concern is google ads / search. Especially as I work with ads.”

These statements reflect a growing sentiment among some users who feel that Google might benefit from fresh leadership.

How about you? Do you think Pichai is still the right person to lead Google?

Google’s Big Push for AI

Despite the criticism, Google is moving full steam ahead with AI.

The company recently expanded AI-powered overviews in more countries, enhancing search with AI-generated summaries and insights.

Google Services also saw substantial growth, reaching $30.9 billion, up from $23.9 billion the previous year, while Google Cloud hit $1.95 billion.

Furthermore, Google’s large language model, Gemini, is being integrated into a wider range of products, providing advanced features across multiple applications.

Google is actively integrating AI across its services, with several experimental products like NotebookLM, Illuminate, and other experimental tools available on AI Test Kitchen.

This heavy investment shows that the tech giant is betting big on AI, perhaps as a way to reinforce its position as a tech leader in a world where AI is becoming increasingly central to innovation.

Final Thoughts

As a developer and AI enthusiast, I agree with the engineers at Google who expressed skepticism about the “>25% AI-generated code” claim.

AI is incredible at auto-completing code and handling repetitive tasks — there’s no denying that. But when it comes to engineering complex systems from scratch or making high-level design choices, AI just isn’t there yet.

The issue isn’t about speed or productivity—it’s about depth.

Right now, AI has limits. It’s still prone to hallucinations, sometimes spitting out code that “sounds” right but doesn’t actually work. For simple tasks or filling in boilerplate code, AI is a fantastic tool. But for projects requiring deep project understanding, reasoning, critical thinking, and problem-solving, AI still falls short.

From what Google’s engineers have shared, this feels more like an assistive partnership than true autonomous coding.

When it comes to Google’s search and advertising services, I believe they still have enough momentum to stay dominant for at least a few more years.

‍