Google CEO Sundar Pichai acknowledged the company is “a bit behind” the frontier on agentic coding.
Pichai called coding “very foundational” to Google’s AI work. He made the comments on the New York Times Hard Fork podcast. The interview came days after Google’s I/O developer conference.
Where Google Sees The Gap
When asked about Google’s position in the AI race, Pichai highlighted areas of strength and those where Google trails.
Google’s models are “very capable” on text, multimodality, voice, audio, and reasoning, he said. But in agentic coding, tool use, instruction following, and long-horizon tasks, Google is “a bit behind at this moment.”
On what that looks like for developers, Pichai drew a clearer line. Google has been strong at creating single-shot web front ends. The gap is in longer-running tasks, where developers work on complex codebases.
Pichai said:
“There is a gap to the frontier where others are, but we are working, you know, we are well aware of it.”
Why Pichai Says Google Lacked Coding Data
Pichai pointed to a developer product gap. Google didn’t have the same external coding product surface generating developer data flows, he said.
He cited Anthropic’s relationship with Cursor as an example. Google “maybe quite didn’t have the surface” that competitors had, he added.
That’s now changing. At I/O, Google announced Antigravity 2.0 as a standalone desktop application for agent-based coding workflows. Internal usage at Google has been growing fast, according to Pichai.
He added:
“We are doubling every week and people are really putting the models to work. That is helping us hill climb quite a bit.”
At the I/O keynote, Pichai had shared internal token usage numbers. He called the growth unlike anything he’d seen inside the company.
What Pichai Said About Gemini 3.5 Flash
The interview came a day after Google launched Gemini 3.5 Flash and made it the default model for AI Mode globally. Pichai acknowledged early complaints about pricing, model quality, and usage limits.
Google had tightened usage limits at launch to avoid outages, he explained, calling the restrictions “rightfully a source of frustration.” The company would make progress on limits “very soon.”
On model quality, he acknowledged that the new model could have regressions in some areas. Some issues are “easy to address” through post-training, and Google would fix them quickly, he added.
SEJ covered the Gemini 3.5 Flash launch and other I/O announcements earlier this week.
Why This Matters
Pichai’s comments go further than Google’s I/O keynote messaging on where the company trails. At the event, the focus on Gemini 3.5 Flash and Antigravity was more confident. This interview offered a more candid read on the competitive picture.
Pichai described the gap as a feedback loop problem. Coding products that developers use daily generates interaction data that improves the next model. By his own account, Google is now building that loop through Antigravity after lacking a similar developer-facing surface.
Looking Ahead
Pichai said Google is making progress on coding and called the space “very dynamic.” Google says Gemini 3.5 Pro is being used internally and is expected to roll out next month. It hasn’t said whether the model will narrow the coding gap Pichai described.
Featured Image: YouTube / Hard Fork podcast
