We turn high signal in-person events for the top AI engineers, founders, leaders, and researchers in the world into the best free learning opportunities for millions around the world here on YouTube. Your subscribes, likes, comments, speaking, attendance, or sponsorships goes a long way toward making our biz model sustainable indefinitely. We strongly believe this industry deserves a better class of community and that we know how to do this well; we just need your support.
Sleeper agents in fine-tuned large language models present a critical security risk, as backdoors remain dormant during standard behavioral evaluations and only activate upon specific, benign triggers. Traditional monitoring, including joint feature analysis, fails to isolate these malicious signals because they are di...
AI engineering is transitioning from a "factory model" of fixed, pre-determined harnesses to an "adaptive engineering" paradigm. While current tools like CLI coding agents and IDEs rely on rigid, upfront configurations to ensure reliability, these systems struggle with the messiness of real-world, multi-agent, and cros...
The "knowledge gap" between rapidly evolving LLM reasoning capabilities and stagnant search retrieval systems limits the effectiveness of AI agents in complex tasks. While models are highly capable, they often rely on inefficient, keyword-heavy queries optimized for legacy benchmarks, leading to significant performance...
The performance of AI agents depends more on the "harness"—the surrounding framework and tools—than the underlying model. By prioritizing robust harness design, developers can achieve high-level performance using local, open-source models, reducing dependency on proprietary systems. Effective agent development requires...
Building AI systems requires prioritizing human discernment over passive approval to combat cognitive surrender, where users uncritically adopt AI outputs. Research, including a case study on the Duolingo English Test, demonstrates that even skilled reviewers often defer to AI signals, effectively rubber-stamping error...
Fable, the latest model from Anthropic, represents a significant shift toward agentic coding where the conceptual "map" of a project meets the complex "territory" of a real-world codebase. Successfully navigating this transition requires "unhobbling" the model by recognizing its spiky capabilities, such as using code e...
Verifiable Continual Learning (VCL) enables AI agents to improve from experience without forgetting previous successes or introducing regressions. This approach shifts focus from simple model fine-tuning to a multi-layered strategy involving harness and memory updates. Because raw production logs lack the structure for...
MCP Apps represent the next evolution of the Model Context Protocol, shifting from simple JSON-based tool calls to interactive, UI-rich experiences embedded directly within LLM chat interfaces. By leveraging sandboxed iframes, these applications enable bidirectional communication between the user and the host, allowing...
The AI Engineer World's Fair centers on the transition of software development from manual coding to autonomous "software factories" driven by agentic loops. This paradigm shift emphasizes orchestration over individual code generation, where engineers design and manage complex, self-improving systems that handle the fu...
"Skill Hell" arises when developers struggle to integrate freely available AI skills, leading to poor performance and organizational inefficiency. To overcome this, evaluate skills using a four-part framework: trigger, structure, steering, and pruning. First, determine if a skill is user-invoked or model-invoked to bal...
Follow this podcast in Podwise
Sign in to get AI summaries, transcripts and mind maps for any episode, including new ones.
