fast.ai founder Jeremy Howard critiques the 'vibe coding' illusion, arguing that AI-assisted tools create a slot machine-like experience that erodes true software engineering skills. He revisits the origins of ULMFiT, champions interactive programming for building intuition, and reframes AI risk from existential threats to the dangers of power centralization and human enfeeblement.
Sherwin Wu, head of engineering for OpenAI’s API platform, discusses the radical transformation of software engineering. He shares how 95% of OpenAI engineers use Codex to manage fleets of AI agents, cutting code review times from 15 to 3 minutes. Wu explores the widening productivity gap, the changing role of managers in an AI-first world, and why "models will eat your scaffolding for breakfast", urging developers to build for where AI is going, not where it is today.
Sam Altman discusses the future of AI, covering the evolution of software engineering, the challenges for AI startups, the roadmap for model capabilities and costs, and the broader societal impacts on economics, security, and education.
A deep dive into orchestrating swarms of AI agents for large-scale code refactoring. Learn practical workflows, task decomposition strategies, and context-sharing patterns, demonstrated with a real-world case study on migrating a large codebase.
Alexander Embiricos, product lead for OpenAI's Codex, discusses the vision of AI as a proactive software engineering teammate, not just a tool. He covers the product decisions that led to Codex's 20x growth, how it enabled shipping the Sora Android app in 18 days, and why the real bottleneck to AGI-level productivity is shifting from model capability to human review speed and interaction.
This talk explores the history of high-assurance software, from the near-perfect code of the NASA Space Shuttle to the Airbus A320. It argues that while these rigorous engineering and formal verification practices were once too expensive for mainstream use, the economic shift brought by AI and agentic coding can make aerospace-level reliability a practical reality for a wide range of applications, solving the key quality limitations of current AI-generated code.