AI Talks

The Agent In Your Apps

The buzzword of the quarter? Agents. These are AIs that don’t just answer questions but take initiative, string tasks together, and act on your behalf. Amazon launched an agent for its sellers, Figma made its design tools more agent-friendly, and banks increasingly leveraged agents for fraud prevention. Even enterprise software like Excel is integrating Copilot functions with agent-like behavior.

Startups were quick to follow: ex-Google engineers debuted Asimov, while others built AI-powered versions of Excel and audio research apps like Huxe. By September, agents were no longer mere experiments—they were rapidly becoming a default interface. Thomson Reuters even outlined “pillars” for professional-grade agentic AI, signaling that this shift transcends consumer novelties, fundamentally reshaping serious industries.

Governments in the AI Race

If Q2 was about flashy product launches, Q3 sharply focused on geopolitics. Washington unveiled a national AI Action Plan (though critics noted its lack of detailed specifics), Canada committed to a B2B strategy, and Shanghai poured billions into subsidies. The U.S. also launched USAi, a secure AI platform designed for federal employees.

China, meanwhile, made headlines with DeepSeek’s open-source models and a push for a global AI consensus. The quarter also revived debate over the talent race, with renewed focus on Song-Chun Zhu’s 2020 move to China, a powerful reminder that AI dominance hinges as much on human talent as on technological innovation.

And then there were the rules: the EU continued pressing ahead with robust regulation, while in the U.S., a proposed moratorium on AI development collapsed under political pressure. Across the board, governments discussed control and oversight, but private companies often moved at a far swifter pace.

Energy: The Hidden Battle

You can’t talk about AI without talking about power—specifically, electricity. Studies continued to highlight the staggering energy cost of a single AI prompt, and analysts repeatedly warned of data centers straining national grids. Tesla even reportedly scaled back its Dojo supercomputer operations, citing prohibitive costs.

Paradoxically, the energy industry is increasingly relying on AI demand for growth, positioning power plants as essential lifelines for data centers. The geopolitical struggle for AI supremacy is now overtly a fight over kilowatts as much as code.

Everyday AI: From Stethoscopes to Movies

Away from the policy debates, AI continued to show up in unexpected places. Doctors unveiled a 15-second AI stethoscope that can detect major heart conditions. Hollywood grappled with AI’s expanding role in filmmaking—while OpenAI reportedly ventured into creating a feature-length animated movie itself.

Finance was another hotspot: AI models passed all three levels of the CFA exam in minutes, CFOs automated routine tasks, and retail investors remained divided over trusting AI for financial advice. In parallel, surveys revealed that nearly all marketing teams now rely on AI tools, indicating that creative departments are being transformed by automation just as profoundly as finance.

On the consumer front, Google Meet enhanced its offerings with real-time translation, NotebookLM gained advanced study tools, and ChatGPT began testing “Pulse,” a proactive mode where it takes initiative. These developments, though less flashy than a super-agent, demonstrate how rapidly AI is permeating daily routines.

Human Unease: Jobs, Trust, and Control

The unease surrounding AI was palpable this quarter. AI pioneer Geoffrey Hinton reiterated his warnings of mass unemployment. Polls showed a majority of Americans feared AI would eliminate jobs—even as another survey revealed their strong desire to keep AI out of their personal lives.

Doctors expressed concerns about over-reliance on AI for critical diagnoses. Investors fretted over a potential AI stock bubble, questioning the boom’s sustainability. And in a very 2025 twist, the language used in Silicon Valley to describe AI took on increasingly quasi-religious tones.

The Future: “World Models” and Super-Agents

Amid the ongoing developments, researchers offered glimpses into what’s next. MIT experts suggested that the future of AI lies not in today’s conversational chatbots but in “world models“—systems capable of learning autonomously by simulating reality. Google’s Gemini and other cutting-edge models are already inching toward this capability.

Combine that with the rise of “super-agents” predicted earlier this summer—AIs designed to manage other AIs—and a clearer picture emerges of where things are heading: systems that function not merely as tools but as full ecosystems, capable of managing complex tasks, coordinating with people, and perhaps even interacting with each other.

Wrapping Up

Q3 2025 unequivocally showed us a field moving from novelty to necessity. Agents went mainstream, governments scrambled for strategic position, energy emerged as a critical bottleneck, and public trust experienced noticeable shifts. Whether you’re a policymaker, a developer, or simply someone navigating your workday, the question isn’t if AI will affect you—it’s how fast, and on whose terms.

One thing’s certain: if the past three months are any indication, Q4 won’t be quiet.

---

This post was researched and written with the assistance of various AI-based tools.