Firehose

This video demonstrates a structured workflow for building a custom internal CRM using Claude Code and a set of pre-defined project templates. The creator emphasizes "shaping" the application by explicitly defining scope and data models before generating code, focusing on hard-coding specific business logic rather than building complex, flexible interfaces. - Builder Methods Pro — Community and training for building AI-powered apps. - Free Workshop — Training on the creator's app-building methodology. - Free tools for builders — A collection of PRD templates, design systems, and app starters.

This video argues that AI coding tools fail to boost productivity because they are being bolted onto broken, fragmented workflows rather than being used to redesign the entire software development lifecycle. The speaker advocates for shifting from "vibe coding" to spec-driven development and using AI agents to bridge communication gaps between product, QA, and operations teams.

This talk argues that most SaaS growth plateaus are misdiagnosed as messaging or acquisition issues when they are actually failures of customer understanding. The speaker advocates for replacing shallow funnel metrics with a customer-led growth framework that maps the emotional journey and specific value moments of the user.

This video is a showcase of long-running AI agents using Google's Agent Development Kit to complete multi-day software engineering tasks. Addy Osmani demonstrates how these agents can autonomously iterate through complex pipelines—such as optimizing 3D assets for the web or building a functional desktop OS—without human intervention.

This video explains how to connect AI agents to enterprise data by using the Looker MCP Server. It demonstrates how to use the Model Context Protocol to expose Looker's semantic layer as a set of tools, allowing agents to query data through governed business logic rather than raw SQL.

This is a candid, conversational panel from the a16z Show featuring Marc Andreessen and Ben Horowitz discussing why founders should abandon traditional media training in favor of direct, authentic communication. They argue that "new media" rewards polarizing, expert-led storytelling over the buttoned-up, risk-averse approach typical of legacy press.

This is a roundtable discussion where three AI industry professionals contextualize recent headlines, specifically debating whether a Cloudflare report on bot traffic signals a structural change in web usage or simply reflects the evolution of search. The panel also briefly touches on Microsoft’s latest model releases and the emergence of AI-generated films at the Tribeca Film Festival.

Addy Osmani outlines three architectural requirements for moving beyond stateless chatbots: event-driven dormancy, durable checkpointing, and separated evaluation. It is a high-level conceptual overview of how to structure persistent, multi-step workflows using the Agent Development Kit.

This video outlines seven workflow patterns for using Antigravity 2.0 to manage coding tasks within existing enterprise systems. The presenter focuses on treating the agent as a junior developer by formalizing team standards into reusable skills, automating test-correction loops, and utilizing background tasks to maintain development velocity.

This is a technical walkthrough of how to use the Agent Development Kit to connect an AI agent to an external tool via the Model Context Protocol. The video explains the MCP handshake process and demonstrates how to wrap a Python function—in this case, a Google Trends fetcher—so an agent can discover and execute it as a tool.

This talk outlines a five-pillar framework for moving AI agents from prototype to production: evaluation, observability, data foundation, orchestration, and governance. The speaker argues that most failures stem from prioritizing model selection over building the measurement and tracing infrastructure required to maintain reliability in real-world environments.

A high-level overview of the architectural patterns required to move AI agents from demos into production. The speaker focuses on treating agents as coordination layers that manage state, enforce business rules, and handle exceptions within existing enterprise workflows. Learn more about their framework here.

This video provides a technical walkthrough of a Kubernetes-based agent that orchestrates a pipeline between a local webcam, Gemini, and Google’s generative media models. You can find the implementation details and source code in the GitHub repo.

This livestream is a walkthrough of a text-based adventure game designed to teach developers how to build autonomous agents using the Agent Development Kit. The presenters demonstrate how to set up a Python-based agent that uses Gemini to parse game state and execute moves, serving as a practical, if niche, introduction to the framework's API and integration patterns.

A practical demonstration of how to use CSS style queries to handle component theming and conditional styling without relying on manual modifier classes. The video shows how to trigger layout changes based on custom properties, effectively letting parent containers dictate the appearance of their children.

A practical test of Figma's native AI agents, demonstrating that they currently struggle to apply design tokens or variables unless a full component-based design system is already connected. The author concludes that while the tool shows promise for mobile flows, it remains less reliable than using Claude Design or Claude Code for actual UI generation.

This is a roundtable discussion from IBM’s *Security Intelligence* podcast where three cybersecurity analysts debate the risks of Apple’s new AI-driven password management features. The panel focuses on the trade-offs between automated security hygiene and the potential for new attack vectors or false positives.

This video provides a walkthrough of a new Agent Development Kit (ADK) designed for building event-driven, long-running AI agents. The presenter demonstrates an HR onboarding workflow to show how the system handles state persistence and multi-agent delegation without the token bloat or continuous polling typical of standard chatbot architectures. You can find the full tutorial here.

This is a conversational interview with a veteran visual designer who shares his experience using Claude to build his first Mac app, Gather. The discussion focuses on his iterative prompting strategy, his shift from Figma-first design to code-first exploration, and his approach to managing technical complexity as a non-developer.

This is a hands-on demonstration of the Agent Development Kit (ADK) and the Agent2Agent (A2A) protocol, showing how multiple specialized AI agents can collaborate on a project while flagging conflicting decisions for human review. It functions as a high-level showcase of how these tools orchestrate workflows across distributed agent nodes.

This presentation outlines three technical strategies for reducing the latency of diffusion models: dynamic quantization, latent caching, and step distillation. The speaker explains how these methods can be combined to achieve real-time generation, with implementation details and tools available in the FastGen repository.

This video explains how Kagenti uses infrastructure-level identity—specifically SPIFFE, Keycloak, and Istio—to prevent "confused deputy" vulnerabilities in multi-agent systems. It demonstrates how to replace static API keys with cryptographic delegation chains to ensure agents only access tools they are explicitly authorized to use.

A conversation between Smitha Kolan and Addy Osmani about the risks of "cognitive surrender" when using AI coding tools. They argue that developers must maintain rigorous code review and testing standards to prevent AI-generated technical debt from compromising software stability.

A demo from Google Cloud Next '26 showcasing GenUI (Agent-to-UI), where a Flutter app uses Gemini to generate and refine images for a latte printer. It serves as a high-level conceptual pitch for how A2UI can dynamically assemble app interfaces based on user input.

This video explains why ChatGPT and similar platforms use a "double iframe" architecture to safely render third-party MCP app UIs. It breaks down the security trade-offs between CSP restrictions, `sandbox` limitations, and origin-based storage, showing how this nested approach isolates apps while maintaining functionality.

This video demonstrates a workflow for building an Apple-style landing page by combining Claude Design for the initial layout, Claude Code for local development, and Higgsfield to generate the product imagery and video assets. The creator walks through the process of using AI to bridge the gap between static mockups and interactive web animations.

This is a recording of a live design session demonstrating a workflow for rapid UI iteration and frontend development. The presenters walk through how they use AI-assisted coding tools to bridge the gap between visual prototyping and production code. - Paper — A design tool used for medium-fidelity UI exploration. - Conductor — A development harness used to manage multiple AI-assisted coding sessions. - Inflight — A tool for managing feedback loops and prototype links. - Dialkit — A utility for adjusting UI spacing and design variables. - Agentation — A tool mentioned for managing AI agent workflows.

This video is a promotional recording for a paid workshop on modernizing UX research. The speaker argues for shifting from static, demographic-heavy personas to "functional personas" that prioritize user goals and tasks, and suggests using LLMs to continuously update these personas as new research data becomes available.

This is a high-level conceptual overview of how AI coding assistants function within a developer's workflow. It explains the "pair programming" metaphor and the benefits of using AI for debugging, testing, and documentation, but it does not demo specific tools or provide technical tutorials.

This is a practical primer on B2B SaaS M&A from an investment banker’s perspective, focusing on why founders should actively manage their exit process rather than waiting for buyers to approach them. It demystifies the buyer landscape—explaining the roles of private equity versus strategic acquirers—and outlines the specific ARR and retention metrics that trigger higher valuations.