Eric Schmidt’s Bold Predictions on AI: Token Windows, Agents, and Text‑to‑Action

Former Google CEO and executive chairman Eric Schmidt recently gave a candid talk at Stanford, which was uploaded to YouTube, explaining why Google is falling behind competitors such as OpenAI.

Massive token context windows

Schmidt highlighted the rapid pace of AI development, noting that every six months he feels the need to update his future‑vision talks. He asked the audience what a “million‑token context window” means, explaining that it allows a prompt of up to one million tokens (or words). He said Gemini aims for ten million tokens, Anthropic has already reached 200,000, and OpenAI likely has similar goals.

AI agents and their role

He defined an AI agent as a system that can perform tasks online on behalf of a user—such as purchasing items—and described it as a large language model with memory capabilities.

Text‑to‑Action concept

Schmidt introduced “Text‑to‑Action” as converting natural language into executable Python code. He cited a hypothetical scenario where, if TikTok were banned, an LLM could be instructed to recreate the platform, gather users and music, and launch it within seconds.

He emphasized that any natural language could eventually be turned into Python, effectively providing each person with an unlimited, obedient programmer.

Programming language outlook

Schmidt noted that Python currently dominates AI development, but a newly released language called Mojo may address some of Python’s limitations, though its survival under Python’s dominance remains uncertain.

Industry observations

When asked why Nvidia is valued at $2 trillion while other companies struggle, Schmidt likened CUDA to a “C language for GPUs,” praising its open‑source ecosystem and the difficulty competitors face in replicating it.

He predicted that within a year we will see larger context windows, more capable agents, and widespread Text‑to‑Action use, which could have an impact greater than current social‑media disruptions.

Controversial remarks and apology

During the Q&A, a student asked why startups like OpenAI and Anthropic appear to outpace Google. Schmidt’s blunt comments about Google’s work‑life balance, remote work, and high‑pressure culture sparked community backlash. He later apologized to the Wall Street Journal, acknowledging his mistake and requesting Stanford to remove the video.

He also criticized Microsoft’s partnership with OpenAI as “foolish,” while noting the competitive pressure from companies such as Apple.

Conclusion

Schmidt’s talk underscored the strategic importance of expanding context windows, developing robust AI agents, and enabling Text‑to‑Action, while also revealing how cultural and managerial attitudes can influence a tech giant’s AI leadership.