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.
Real-time voice agents require a cascaded architecture that mirrors self-driving car systems, specifically separating perception, planning, and control layers to balance intelligence with low latency. Perception relies on a streaming speculative transcriber that layers fast, real-time detection with slower, context-awa...
Transitioning voice AI agents from proof-of-concept to production requires addressing critical failure modes in latency, transcription accuracy, and data collection. Achieving optimal performance involves balancing cost and intelligence by utilizing open-source models like Qwen 3.5 or Gemma 4, which offer superior cont...
Voice agent reliability and safety represent the most significant barriers to scaling conversational AI in production. While voice technology is rapidly advancing, current deployments suffer from a 10% error rate, leading to consequences ranging from minor user frustration to critical safety failures in healthcare and ...
Building a voice-first AI companion requires prioritizing low latency and managing conversational volatility, as users frequently interrupt and switch topics. Maintaining a sub-two-second response time is critical for immersion, necessitating a tiered model architecture that routes high-stakes interactions to frontier ...
Voice agent technology has evolved from constrained, closed-ended systems like early Siri to open-ended conversational models, and finally to full-duplex, real-time speech-to-speech interfaces. While modern speech-to-speech models offer superior naturalness and low latency, they often sacrifice the reasoning and tool-c...
Voice agents are evolving beyond simple conversational interfaces, shifting toward three distinct interaction modes: speech-to-speech, speech-to-action, and event-to-speech. Rather than relying solely on verbal responses, developers can leverage existing software structures to enable voice-driven form filling, creative...
Speech-to-speech models represent the future of human-computer interaction, moving beyond traditional cascaded systems toward natively multimodal, end-to-end architectures. By integrating audio, video, and text into a unified token embedding space, these models enable sophisticated agentic capabilities, including real-...
Building LLM agents is shifting from manual, code-heavy orchestration to streamlined, file-based architectures. Historically, developers managed complex Python loops, JSON schemas, and state tracking to enable agent functionality. The introduction of the Interactions API and remote agents replaces this boilerplate with...
Building effective AI agents requires moving beyond simple "mega prompts" toward a file-system-based architecture that allows agents to manage their own state, tools, and execution paths. Vercel’s internal development journey—from a rudimentary data science assistant to a sophisticated, scalable framework—demonstrates ...
Enterprise AI agents require robust "harnesses" and sophisticated memory management to move beyond simple code generation into effective team collaboration. While AI tools accelerate individual tasks, they often fail to share critical context across distributed teams, creating productivity bottlenecks and code divergen...
Building agentic CLI tools requires a shift from simple prompt-based security to deterministic, layered defense. The PostHog Wizard, an agent that automates software setup, demonstrates that tools with command execution capabilities function as "malware starter packs" if left unchecked. Security must be enforced determ...
Running AI agents reliably in production requires moving beyond simple SDKs toward robust infrastructure that supports durable execution and state management. As agentic systems evolve into persistent, asynchronous entities, they demand capabilities like automatic failure recovery, concurrent session handling, and inte...
Loophole, an open-source adversarial agent framework, enables users to codify personal moral principles into a structured legal system and stress-test them against synthetic case law. By utilizing LLMs to simulate adversarial agents, the system identifies contradictions, loopholes, or instances of overreach, prompting ...
Harness engineering defines the architecture surrounding an AI model, encompassing memory, tools, scaling, and observability. Agents fall into two distinct categories: those used for productivity, such as coding assistants, and those built for specific applications. Effective harness engineering requires moving beyond ...
Tokens in agentic systems are not fungible; instead, assigning them distinct functional roles—advising, grading, and dreaming—significantly improves task outcomes. While increasing total token budget generally enhances performance, specific strategies yield superior "alpha" for complex tasks like financial analysis. Ad...
Building ambitious software in the age of AI requires balancing high-velocity development with rigorous architectural standards. Dioxus, a cross-platform framework written in Rust, demonstrates that while AI coding agents excel at handling complex technical knowledge, debugging, and mundane tasks, they cannot replace t...
Follow this podcast in Podwise
Sign in to get AI summaries, transcripts and mind maps for any episode, including new ones.
