After the Google I/O conference, Google CEO Sundar Pichai appeared on the popular tech podcast "Hard Fork" to have an in-depth conversation on core issues such as competition in large models, transformation in search patterns, and AI safety. This rare and candid exchange came amid multiple challenges Google faces in the AI race.

Key Takeaways

  • Model Evaluation: Google is at the forefront of the industry in overall intelligence and multimodal capabilities, but it currently lags behind competitors in agentic coding and handling complex, long-term tasks.

  • Search Transformation: Google will not aggressively switch entirely to an AI model. Search links and sources will remain for a long time, and the business model combining subscriptions and advertising will remain solid in the AI era.

  • Intelligent Agent Promotion: AI intelligent agents (such as the upcoming Spark) will follow a "step-by-step" strategy, with a focus on preventing boundary security and overstepping behavior to avoid users developing a "retreat mentality."

  • Compute Logic: Google continues to open TPU access to competitors, aiming to maintain its leading position in hardware through economies of scale and ecological collaboration.

  • AGI Warning: With the explosive technological progress in recent years, the expected timeline for achieving AGI is accelerating. Even if it is not reached in the short term, technological capabilities will see a generational leap in the next three years, and society must prepare in advance.

Interview Highlights: On Competition, Anxiety, and the Future

1. Honest Admission on "Lagging Behind": Acknowledging the Programming Gap

In response to external criticism, Pichai openly stated, "Our models are highly competitive in text, audio, and general reasoning, but we are slightly behind in agentic coding and following instructions." He described programming as a "core frontier" that Google must hold onto and revealed that the team is rapidly addressing performance degradation issues in Gemini 3.5 Flash through post-training. He believes that "30 to 60 days now feel like five years," and the competitive landscape is changing rapidly.

2. On Search: Not "Removing the Band-Aid"

When asked about when to remove the classic search interface, Pichai emphasized the importance of "guiding users." He said Google will take a gradual and careful approach rather than a radical shift. He pointed out that users still need search links as an efficient way to connect, which is not just a product habit, but also a necessity for information verification.

3. On Anxiety: The Inevitable Pain of Technological Change

When discussing public resistance to AI, Pichai showed a manager's prudence. He believed that human evolution cannot fully keep up with the astonishing speed of technological advancement, and anxiety is "natural and reasonable." However, he pointed out that the positive impacts of AI (such as improved efficiency in medical radiology) are underestimated, and he does not agree with overly pessimistic predictions. But he also emphasized that society must participate and express its preferences to guide technology toward the right direction.

4. On AGI: At a Critical Stage of Evolution

Regarding whether the "singularity" has arrived, Pichai said that moving toward AGI is an inevitable trend. Although he avoided giving a specific date, he clearly admitted that technological progress in the past one or two years has exceeded his original expectations, and the realization of AGI is closer than anticipated. He emphasized, "The public should not relax their preparation simply because they believe AGI is still far away."

Market Interpretation: Google’s Conservative Stance and Initiative

Pichai’s interview sent a clear signal: Google is transitioning from being a "versatile player" to focusing on differentiated competition.

  • On the Hardware Side: Google maintains its role as an AI computing "army supplier," opening TPU access to control the pace of technological iteration.

  • On the Product Side: Faced with the early advantage of Anthropic and OpenAI in programming tools (such as Cursor integration), Google is showing strong reflection and a drive to catch up.

  • On Trust: Google's "transparency" strategy on intelligent agent products indicates its deep understanding that the biggest enemy of AI implementation is not technical bottlenecks, but users' "lack of trust."

This interview is not only a response to external doubts, but also a reminder to the entire industry: in this profound AI transformation, no company can always stay in a leading position. Only through continuous technological iteration and deep social responses can companies remain unbeaten in the next three years.