Intercom doubled its engineering throughput in under a year by treating its AI coding agent not as a simple tool, but as a new senior engineer. This involved a full onboarding process onto their 15-year-old Rails monolith, creating a library of durable skills for recurring tasks, and providing audited access to all internal systems.
Andrej Karpathy discusses the shift from 'vibe coding' to 'agentic engineering,' explaining why LLMs should be treated as 'ghosts'—jagged, statistical entities—rather than animals. He delves into the Software 3.0 paradigm, the limits of verifiability, and why human understanding remains the ultimate bottleneck in an age of outsourced thinking.
An analysis of the current unstable equilibrium between open and closed AI models, arguing that closed models will likely pull ahead due to economic and data feedback advantages. The long-term, stable future for open models lies in a specialized ecosystem of cheaper, faster models, potentially funded by new structures like consortiums.
James Shaw, a mechanical engineer and Ansys channel partner, delves into the current and future impact of agentic AI and physics-informed neural networks (PINs) on simulation workflows. He explores how AI is revolutionizing aspects from tech support and model setup to the solver itself, particularly in CFD. The discussion also covers the implications for the engineering job market, the 'senior-junior inversion crisis', and the continued irreplaceability of skilled engineers due to the inherent physicality of the world, emphasizing the need for robust, trustworthy data to train AI.
Variance, emerging from three years in stealth with a $21 million Series A, is transforming enterprise risk and compliance through purpose-built AI agents. Founded by ex-Apple engineers, the company automates complex tasks like fraud detection, content review, and identity verification for Fortune 500s and platforms such as GoFundMe. They discuss the strategic reasons for stealth, technical challenges of integrating disparate data sources (including UI scraping), the shift from legacy systems to self-healing AI agent architectures, and how their lean, AI-maximalist team detects sophisticated threats like state-sponsored fraud rings.
François Chollet discusses his contrarian approach to AI, moving beyond scaling LLMs to understand intelligence from first principles. He explains his work on the ARC benchmark series, including the new ARC-AGI V3, designed to measure 'agentic intelligence' and skill acquisition efficiency. He also introduces his lab, Ndea, which is developing a new ML paradigm based on symbolic models, and shares his perspective on the limits of current systems and the future path to AGI.