AI is making advances on multiple fronts at once. Looking beyond individual announcements, we identify which ones signal meaningful change, what remains uncertain, and where today’s progress may take AI next.

The capabilities of AI are advancing faster than our understanding of the technology itself. In hope of keeping pace, researchers are racing to make models more interpretable and better aligned.

The next acceleration in AI may come from AI itself. Researchers are already using AI to design, test, and improve new systems. We’ll ask where that path leads as more of the research process becomes automated.

AI is getting better at representing the physical world, opening new possibilities for robots, autonomous systems, and physical AI. The challenge is building systems that can model environments, anticipate consequences, and act within them.

As AI improves, measuring progress is becoming harder. Familiar benchmarks are losing their usefulness, forcing researchers to develop new forms of evaluation and better ways to tell when a model is genuinely getting better.

Today’s AI can look astonishingly capable in one moment and unexpectedly brittle in the next. We’ll examine what drives that unevenness, how well models generalize beyond familiar patterns, and what their failures reveal about the shape and limits of machine intelligence.

As AI takes on longer and more consequential tasks, consistency becomes as important as capability. This session explores how reliability changes over time, where models still become uncertain or unstable, and what dependable performance really requires.

The biggest gains from AI may come from changing work itself. Rebuilding workflows around new capabilities means defining the right boundaries between people and machines rather than simply bolting new tools onto old processes.

Moving AI into production means building the system around the model. Access, integrations, orchestration, guardrails, and human oversight all must work together before AI can operate reliably at scale.

What happens when your customer is an AI agent? A web increasingly shaped by machines acting on behalf of people could transform discovery, commerce, customer relationships, and the way organizations establish a digital presence. 

When it comes to cybersecurity, agentic AI expands both the attack surface and the blast radius. Prompt injection, compromised tools, and other vulnerabilities become more consequential when swarms of autonomous agents can discover and exploit them at unprecedented scale. 

AI is changing the rules of business. We examine which assumptions are breaking down, where expectations run ahead of evidence, and what the emerging reality means for how companies compete, create value, and grow.

AI is reducing the time required for research, analysis, and other high-value professional work. When efficiency begins to undermine the pricing model, firms have to rethink fees, productivity, and the economics of expertise.

The best sales process has always been expensive because it depends on time, research, and personal attention. AI changes that equation by making individualized selling possible at far greater scale, with potentially profound implications for how companies reach customers.

Even before AI changes our circumstances, uncertainty about its impact can reshape our lives. We examine the cost of anticipating disruption to careers, security, and identity and explore how to preserve agency while preparing for an unsettled future.

AI makes answers readily available, but understanding still takes work. What happens when we bypass that work? We explore how human expertise and judgment develop and what we risk losing when the process of thinking becomes optional.

As AI becomes more powerful and more embedded in our lives, who ultimately controls it? Power is increasingly distributed across model makers, compute and cloud providers, capital, governments, regulators, and open ecosystems with major implications for competition, choice, and accountability.