“Developer loyalty is at zero right now”: Google doesn’t care which AI coding tool you use

“Developer loyalty is at zero right now,” Richard Seroter, Google Cloud’s Senior Director and Chief Evangelist, told The New Stack in a frank interview at the Google Cloud Next conference last week. For a company that offers many developer tools — and maybe too many — that’s a tough position to be in. But Seroter isn’t worried about it.

“Our job is that we should make it easy. If you want to build with AI or if you build AI apps, we should be excellent at both and be the best at both. We are not ceding anything. We will be the best at AI dev tools. We’re gonna get there — always. I mean, that’s everybody’s goal. We’re gonna keep pushing there,” Seroter said.

“Developer loyalty is at zero right now.”

But Google is also pragmatic there. Even if a developer isn’t using Google’s developer tools, it still wants to be the platform where the code runs.

That also extends to the experience of using models — and anyone’s models — on Google Cloud, he argues. “If you look at the latency, the best performing way to use Anthropic is on Vertex over Azure, AWS, and Anthropic itself. And so even if you’re using Claude Code, you should be using Vertex — otherwise you’re just getting worse performance.”

“[Google Cloud CEO] Thomas Kurian told me a while back, when we started building our agent skills, we should have the best dev experience for anyone who’s deploying to Google Cloud. I don’t care what tool you use. Cursor? Unbelievable.

“When you pull in our skills, or you pull in our MCPs or instructions, that should be better than what Amazon offers you,” Seroter said. “And so you get the best advice, you get the most accurate advice. You get the right architecture that’s correct the first time. So if you choose a different upstream toolchain, this should be your default. It’s always gonna be better.”

Google Cloud is a platform, not a holding company

He was clear that Google wants to win in all of these areas, but he also acknowledged that it’s hard to be best-of-breed at everything. Taking a bit of a swipe at the competition, Seroter noted a philosophical difference in how Google Cloud approaches its platform.

“Not everyone agrees with me, but I don’t think we are a portfolio company with 300 products. We’re a platform company with one platform.”

“I don’t think Google sells 300 services. I think we sell one platform,” he said. “Not everyone agrees with me, but I don’t think we are a portfolio company with 300 products. We’re a platform company with one platform. […] I don’t want a holding company in a cloud — and I think you still have those experiences in some clouds.”

Sometimes, he believes, it’s okay for some services to be “good enough,” but in some areas, Google Cloud wants to be better.

“We’re going to pour ourselves into certain types of services that have to be excellent,” he said. “So yeah, we can do more because we’re a bigger team, but I think we all are selectively ruthless where we need to be more focused. And that’s probably been the biggest change company-wise in the last three years — is that we’ve just decided to be laser focused in a few areas, especially around AI development and excellence that I love — and that comes from the top.”

Is Google not moving fast enough?

Google has been criticized a bit lately, especially by former Google engineer Steve Yegge, who argues that Google isn’t moving fast enough, in part because its own use of AI in engineering isn’t up to par and because engineers are not able to — or don’t want to — use the best available tools.

In a recent post, Yegge specifically mentioned that his sources told him many of Google’s engineers aren’t allowed to use Claude Code “because it’s the enemy,” even as the DeepMind team uses it.

Seroter, unsurprisingly, doesn’t agree and argues that Google, given its outsized importance in the ecosystem, has to strike a balance between moving fast and keeping things stable.

“They’re usually working with incomplete information. I know all the parties involved,” he said of the recent criticism. “Steve’s work? I reviewed his Vibe Coding [book] as a tech reviewer. [Co-author] Gene Kim sent me a copy. But you’re working with incomplete information. Google is a big company.

“Talk to his buddy, the engineering manager here, who says this isn’t working, but it’s different than here and here. So, formally, are we doing it all the same way? Of course not. Android operates differently from Cloud and DeepMind, but at the same time, the push is to ensure we never sacrifice quality.

“We’re gonna do it fast, but we’re gonna do it safely.”

“So as much as it would be cool to vibe code the next extension to [Google] Maps, I can’t break Maps. We don’t have the luxury of shipping a chatbot. We have the luxury of running the most important platform and infrastructure on the planet. So we’re gonna do it fast, but we’re gonna do it safely. I don’t think everyone has that obligation, and so there’s a certain freedom to be a little more YOLO in some companies than ours. Doesn’t mean we’re still not going to be aggressive and fast — but I can’t break the internet.”

Seroter also believes that Google is moving “pretty darn fast” and pointed at the new TPU hardware, agent platform, and model updates. “It is an amazing portfolio. How much faster do I need to go? I don’t know, but our customers can not keep up with any of this stuff. No one does. How much faster do I need to be? Is speed the entire thing? Or is that still a way to ship quality products to customers? I think sometimes we lose sight of that, industry-wise, and it’s purely about the speed. Speed to what, like speed, without direction is worthless.”