Moving an AI PoC from 50% to 100% reliability requires a new development paradigm. This talk introduces a practical, evaluations-first approach, reverse-engineering tests from real-world user scenarios and business outcomes to build a robust benchmark, prevent regressions, and enable confident optimization.
Sarah Goa from the AI-native venture fund Conviction discusses the state of AI, arguing that despite the hype, building valuable AI products is challenging. She covers the rapid advancement in AI capabilities like reasoning and multimodality, the increasingly competitive model market, and provides a playbook for building successful applications, using "Cursor for X" as a framework. The key takeaway is that defensibility in AI comes from superior execution and deep workflow integration, not just the underlying model.
Danielle from the Amazon AGI SF Lab introduces a new paradigm for agent development called "Useful General Intelligence" (UGI). Instead of focusing on making AI smarter than humans, UGI aims to build AI that augments human intelligence and agency. This talk explores the cognitive science principles behind this vision and introduces Nova Act, an agentic model and SDK designed to reliably interact with computer UIs, turning the browser into a programmable tool.
A deep dive into the design principles for building effective AI agents, covering the evolution of the ReAct loop, the critical role of memory and learning from experience, the 'build vs. buy' dilemma for tooling, and the importance of abstracting all capabilities—including systems and people—as tools.
Fal co-founder Burkay Gur and head of engineering Batuhan Taskaya discuss their journey building a high-performance generative media cloud. They cover their strategic pivot to media models, core optimization principles born from early GPU scarcity, and the development of a customer-obsessed culture to navigate the fast-paced AI model landscape.
Dr. Jasper Zhang argues that the relentless construction of new data centers is an inefficient, expensive, and unsustainable solution to the AI compute demand. He proposes a global GPU marketplace as a superior model, designed to aggregate fragmented, idle resources, drastically reduce costs through efficient allocation, and ultimately democratize access to AI infrastructure for developers and startups.