Henrik Kniberg shares practical experiences from over two years of AI agent development, defining agents as autonomous entities with missions, tools, and an LLM brain. He covers effective design patterns, architectural insights, and safety considerations, emphasizing the importance of human-in-the-loop collaboration and iterative development to build agents that are not just powerful, but genuinely useful in real-world scenarios.
This episode of Security Intelligence explores the "Promptware" kill chain for AI attacks, moving beyond simple prompt injections. The discussion also covers evolving cloud attack trends targeting ecosystems over infrastructure, ransomware gangs "living off the land" with native tools, and the critical yet overlooked "rusting edge" of OT security.
Professor Heather Kulik shares her hard-won perspective on applying AI to materials science, from discovering novel polymers with surprising quantum properties to the practical limitations of LLMs and the critical need for integrating deep domain expertise with data-driven methods.
Dexter Horthy of HumanLayer critiques the initial Research-Plan-Implement (RPI) framework for AI coding agents, revealing its tendency to encourage 'outsourcing thinking'. He introduces CRISPR, a new structured methodology that emphasizes smaller, focused prompts, human-agent alignment through artifacts like Design Discussions, and engineer ownership to combat 'slop' and improve code quality in complex projects.
Scott Breitenother, CEO of Kilo, discusses the evolution of software development, where engineers are shifting from writing code to orchestrating AI agents. He shares lessons from processing 25 trillion tokens, emphasizing the critical role of trust, the importance of end-to-end ownership, and how this new paradigm leads to a 10x increase in shipping velocity.
Professor Kyunghyun Cho, a co-author of the first paper on attention, discusses the future of AI. He argues that today’s models have already captured most correlations in passive data, making the real challenge about actively choosing which data to collect. He also explores the open debate around world models, the surprising lack of coding agent adoption among his students, and the foundational work that led to Retrieval-Augmented Generation (RAG).