Gemini 3 has topped nearly all benchmarks, and is one of the best models available for coding, but just six months ago, it wasn’t even being used at Google itself.
That’s according to Sergey Brin, Google’s co-founder and one of the most influential figures in tech, who returned to a more active role at the company in recent quarters. Speaking at the All-in Summit in Miami in May this year, Brin had revealed a startling internal conflict that exposed the bureaucratic friction slowing down even the world’s most advanced AI company. His admission—delivered with visible frustration—offers a rare glimpse into the organizational challenges that can emerge when a company’s internal policies fail to keep pace with its own technological breakthroughs.
“I’m embarrassed to say this,” Brin began. “Recently, I just had a big tiff inside the company because we have this list of what you’re allowed to use to code and what you’re not allowed to use to code. And Gemini was on the no list, for a bunch of really weird reasons. That boggled my mind,” he said.
The term “vibe coding” refers to the increasingly popular practice of using AI tools conversationally to generate, debug, and iterate on code—a workflow that has become central to modern software development. That Google’s own flagship AI model was banned from this use case internally seemed almost incomprehensible to Brin, who had returned to Google after retiring when he saw the developments in AI.
“I mean, nobody would enforce this rule,” Brin continued. “But there was this actual internal webpage, for whatever reason, historical reason, somebody had put this and I had a big fight with them. I cleared it up after a shockingly long—I definitely told Sundar [Pichai] about it. He was very supportive. It was more like I was like, I talked to him, I was like, I can’t deal with these people. You need to deal with this. I’m beside myself that they’re saying we can’t vibe code using Gemini.”
Brin’s exasperation was palpable, and his decision to escalate the matter to CEO Sundar Pichai underscores how deeply the issue troubled him. “It did get fixed and people are using it now,” he confirmed. “We’re trying to roll out every kind of AI, and also testing out the Cursors of the world, all of those, to just see what makes people more productive.”
As it might’ve been little pushes that these from its co-founder that have led to Google’s big AI turnaround. Until last year, Google was widely seen to be lagging in the AI race, with no top tier models, a botched release of image generation abilities, and no major consumer platforms. Things have turned around in the last few months though. Google now has the best AI model by a distance in Gemini 3, it has the best image generation models in Nano Banana Pro, and has seen increasing traction on Gemini’s consumer-facing app. And while there might’ve been many factors that led to the turnaround, the return of the co-founder who questioned — and fixed — things like Google’s own AI models not being used at the company must’ve played a role in the course correction.