Source Library

Curated YouTube notes (Nate B Jones and Every) backing the Marq AI OS argument and the Q3 OKR. Each entry has a one-line blurb and an Obsidian link to the local copy in 00-raw-sources/. Originals live in the main vault at vault/life/inbox/youtube/.

How these map to the OKR

• Binary goals (train the company on the agentic-workflow toolkit; build workflows) — Everyone Is Prompting Better (the scaffolding definition), Claude Code vs Codex, Anthropic’s $2.5B Leak, Claude Code for Absolute Beginners, How We Built Claudie, Vibe Code Camp.
• Behavioral change (measurement, baseline, targeted training) — Everyone is Getting AI Fluency Wrong (10-level framework), Shopify (visibility → org learning), Your Claude Limit Burns in 90 Minutes (context/compaction discipline), Why Your Best Employees Quit.
• Compounding outcomes (systems that generate output; impossible 3x goals) — Compound Engineering Explained, The Trillion Dollar Agentic Workflow Opportunity, The Work Primitive, The AI Sandwich, 20 People $100M Revenue, Why We Switched to Codex (OS-for-knowledge-work framing).
• Dependencies (CLI context tool, context cortex, session sync) — The Real Problem With AI Agents (context gap), Your AI Agent Fails 97.5%, The One AI Writing Hack (agentic data room), The Infrastructure Nightmare.
• Stakeholders / change management (Owen mandate) — Why Your Best Employees Quit, Everyone Is Prompting Better (“Why Leadership Needs This Too”), Shopify.

Nate B Jones (AI News & Strategy Daily)

• Demystifying AI Agent Scaffolding: Prompts, Skills, Plugins, and MCP Connectors Explained — A clear mental model for building agentic workflows by distinguishing prompts, skills, plugins, and connectors so technical and non-technical users alike can design reliable automation. The foundational “what is an agentic workflow and why it unlocks so much” piece.
  Everyone Is Prompting Better. Almost Nobody Is Packaging Work.

• Why AI Harnesses Matter More Than Models: Claude Code vs OpenAI Codex Architectures — Why the harness around a model now matters more than the model itself, and how harness choice creates deep workflow lock-in.
  Claude Code vs Codex_ The Decision That Compounds Every Week You Delay That Nobody Is Talking About

• Architectural Insights from the Anthropic Claude Code Leak for Building Production AI Agents — Backend engineering primitives behind robust production agents: tiered permissions, workflow state management, and reusable design patterns.
  I Broke Down Anthropic’s $2.5 Billion Leak. Your Agent Is Missing 12 Critical Pieces.

• How to Assess and Improve Your AI Fluency Level: A Comprehensive Framework — A model-agnostic 10-level framework for mapping current LLM skills and planning development from basic user to systems thinker.
  Everyone is Getting AI Fluency Wrong—Steal My 10 Level Framework That Exposes the Real AI Skill Gap

• Why Shopify Forces AI Agent Interactions Into Public Channels for Organizational Learning — Making AI workflows visible across teams solves the hidden-knowledge problem and accelerates collective learning.
  Shopify Made 5,938 People Better at AI. Not With Training. By Watching.

• Stop Wasting AI Tokens: Practical Strategies for Efficient Context and Cost Management — Convert files to markdown, cache prompts, and scope context windows to control compute cost without losing quality.
  Your Claude Limit Burns In 90 Minutes Because Of One ChatGPT Habit.

• Strategic Shifts in Enterprise AI: How PE and Hyperscalers Target the Implementation Layer — The value of enterprise AI lives in the implementation layer that wires agentic workflows into business systems, not in the models.
  The Trillion Dollar Agentic Workflow Opportunity Is Here

• The Future of AI Agents: Moving Beyond Computer Use to Semantic Work Primitives — Agents must evolve past screen interactions into semantic work primitives that understand a task’s context, rules, and consequences.
  The Work Primitive_ What Every AI Product Leader Gets Wrong

• Defining Frontier Operations as the Critical Workforce Skill for the Expanding AI Era — “Frontier Operations” as the dynamic skill set for managing workflows and risk at the shifting edge of AI capability (tiny teams beating giants).
  20 People. $100M Revenue. The 5 Operations Behind Every Tiny Team Beating a Giant One.

• The Structural Flaw in AI Agent Deployments and How to Solve the Context Gap — Agents fail because users can’t articulate tacit knowledge; an interviewer agent extracts and structures personal workflows into config files.
  The Real Problem With AI Agents Nobody’s Talking About

• Why AI Agents Fail at Long-Term Jobs Without Human Context and Evaluations — Agents handle isolated tasks but lack long-term memory and org context; evaluation frameworks and human oversight are essential.
  Your AI Agent Fails 97.5% of Real Work. The Fix Isn’t Coding.

• How to Stop AI Hallucinations by Building an Agentic Data Room Before Prompting — Have agents build a structured data room and source inventory before drafting, so they rely on authoritative context instead of inventing facts.
  The One AI Writing Hack Nobody Talks About.

• Bridging the AI Skills Gap: Why Enterprise Adoption Requires Management and Judgment Over Technical Proficiency — Adoption fails when orgs focus on prompting instead of task decomposition and quality judgment; the missing middle layer is management skill.
  Why Your Best Employees Quit Using AI After 3 Weeks (And the 6 Skills That Would Have Saved Them)

• Why Long-Running AI Agent Systems Require Strict Harnesses — A virtual-town agent simulation shows safe autonomy needs strict system harnesses and permission gates, not just model-level alignment.
  Claude’s AI Town Voted Yes On Everything. That’s Not A Good Sign.

• How AI Agents and Automation are Transforming Data Platform Infrastructure — AI coding agents accelerate app development but strain infrastructure, driving the need for multi-agent systems and automated guardrails.
  The Infrastructure Nightmare Nobody Is Talking About

• Physical Infrastructure Constraints and the Industrial Supply Chain Behind AI Vendor Contracts — AI vendors are bound by power, memory, and packaging bottlenecks; procurement should be treated as an industrial supply contract.
  Why the AI boom is about to hit a wall

Every

• Lessons from Teaching Claude Code to Non-Technical Users — Mike Taylor on adoption barriers, the paradigm shift to agentic AI, context management as the core skill, and why live instruction beats self-paced. The training playbook.
  Claude Code for Absolute Beginners — What We’ve Learned Teaching AI to Non-Coders

• Building an Autonomous AI Project Manager to Automate Workflows with Claude Code — Natalia’s full walkthrough of building an autonomous agent (“Claudie”) for client onboarding and project tracking. Concrete worked example of a built workflow.
  How We Built ‘Claudie,’ Our AI Project Manager (Full Walkthrough)

• A Guide to Compound Engineering and the Future of AI Software Development — The compound-engineering methodology: agents automate workflows and continuously improve the system itself. Cleanest articulation of “systems that generate output.”
  Compound Engineering Explained

• Using AI Coding Agents Like Codex as an Operating System for Knowledge Work — How to turn a coding agent into a centralized operating system for managing complex knowledge work. Direct anchor for the “AI OS” framing.
  Why We Switched From Claude Code to Codex

• The AI Sandwich: How Humans and AI Collaborate in Modern Engineering Workflows — Humans own ideation and final polish; AI owns routine execution, elevating human roles to creative and strategic direction.
  The AI Sandwich_ Where Humans Excel in an AI World

• Building AI-Durable Startups and Validating Ideas with Custom AI Agents — Bolton & Watt’s “slow incubator” philosophy and workflows for using custom agents to validate ideas and automate market research.
  Meet the Slowest Startup Incubator in the World—Pumping Out Billion-dollar Companies

• The AI Paradox: Why Automating Knowledge Work Increases Demand for Skilled Human Experts — Commoditizing basic skills floods orgs with generic work that needs advanced human oversight to refine and implement.
  AI Was Supposed to Save Time. Why Am I Busier_

• Vibe Code Camp: Live Demos of AI Coding Workflows and Agent-Native Apps — Top builders demonstrate personal AI coding workflows and agent-native architectures to ship production apps fast.
  Vibe Code Camp_ Live Marathon With the World’s Best AI Builders

• Lessons From Vibe Coding an App: The Pirate and Architect Engineering Model — Balancing rapid prototyping against structured architecture; experienced engineers remain critical for turning prototypes into production.
  Why Every AI Team Needs Pirates and Architects