Just give me the meat in the sandwich.

Nate B. Jones Digest

Nate B. Jones

You Don't Need SaaS. The $0.10 System That Replaced My AI Workflow (45 Min No-Code Build)

Published: 03 Mar 2026 · 02:00 AM AEDT

Abstract

What's really happening when Claude's memory doesn't know what you told ChatGPT and your phone app doesn't share context with your coding agent? The common story is that AI memory is getting better—but the reality is more interesting when every platform has built a walled garden designed to create lock-in.

Highlights

  • In this video, I share the inside scoop on why the architecture of agent-readable memory matters more than any individual tool:
    • Why your Notion workspace is beautiful for humans and useless for agents that search by meaning
    • How a Postgres database with vector embeddings runs for 10-30 cents a month
    • What MCP servers enable when one brain connects to every AI you touch
    • Where the compounding advantage lives for people who stop re-explaining themselves For anyone watching the agent revolution go mainstream, the gap between starting from zero and starting with six months of accumulated context is the career gap of this decade.
  • Chapters 00:00 Your AI Agent Doesn't Have a Brain 01:30 Why the Second Brain Guide Needed an Upgrade 03:00 The Memory Problem Hiding in Your Prompting 04:30 Why Context Infrastructure Beats Better Models 06:00 The Walled Garden Problem: Siloed AI Memory 07:30 How Corporate Memory Lock-In Works Against You 09:00 Agents Are Mainstream — and They Need Context Too 10:15 Why Note-Taking Apps Weren't Built for Agents 11:30 The Human Web vs.
  • the Agent Web 13:00 Introducing Open Brain: The Architecture That Fixes This 14:15 Why Postgres Is the Right Foundation 15:15 Vector Embeddings and Semantic Search Explained 16:30 What the System Actually Looks Like in Practice 17:45 Capture and Retrieval: How the Two Sides Work 18:45 The Cost: 10 to 30 Cents a Month 19:30 Person A vs.
  • Person B: The Compounding Advantage 21:00 Why This Is the Career Gap of the Decade 22:00 What You Can Build on Top of Open Brain 23:15 Honest Limitations to Know Before You Start 24:00 The Four Prompts That Run the Full Lifecycle 26:30 Weekly Review: Five Minutes That Compound Forever 27:30 What It Feels Like When the System Works 28:30 The Bigger Lesson: AI Forces Clarity of Thought 29:30 Open Brain vs.

References & Links

Why Every AI Skill You Learned 6 Months Ago Is Already Wrong (And What Is Replacing Them)

Published: 02 Mar 2026 · 06:00 AM AEDT

Abstract

What's really happening when every other workforce skill in history had a finish line but AI doesn't? The common story is that humans need to learn new tools—but the reality is more interesting when you picture capability as an expanding bubble where the surface area keeps growing, not shrinking.

Highlights

  • In this episode, I share the inside scoop on why frontier operations is the first skill that expires quarterly:
    • Why boundary sensing against November's model leaves you standing inside February's bubble
    • How seam design structures clean handoffs between human and agent phases
    • What failure model maintenance looks like when agents fail subtly, not obviously
    • Where leverage calibration becomes the scarcest resource in an agent-rich environment For professionals watching capability accelerate, the person who developed this skill six months sooner doesn't have a head start—they have six months of updated calibration their peers can't replicate.

References & Links

My 10-Year-Old Vibe Codes. She Also Does Math by Hand. Why That's the Only Strategy That Works.

Published: 01 Mar 2026 · 03:00 AM AEDT

Abstract

What's really happening when AI tutors double learning outcomes in controlled studies but college professors report students can no longer read a full chapter? The common story is that AI will either save education or destroy it—but the reality is more interesting when the calculator moment of the 1970s reveals the principle we've forgotten.

Highlights

  • In this video, I share the inside scoop on why foundation before leverage is the only position that makes sense:
    • Why my 10-year-old does long division by hand and vibe codes with Claude in the same week
    • How specification quality determines the gap between agentic success and catastrophe
    • What the metacognition skill of thinking about your own thinking actually looks like in practice
    • Where cognitive offloading quietly erodes capability before anyone notices the muscle weakening For parents watching the world change faster than any curriculum can track, the gift we give our kids is the cognitive architecture that lets them direct intelligence rather than depend on it.

References & Links

'Prompting' Just Split Into 4 Skills. You Only Know One. Here's Why You Need the Other 3 in 2026.

Published: 28 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when two people sit down with the same model on the same Tuesday and one of them produces a week's worth of work before lunch? The common story is that better prompting means better instructions—but the reality is more interesting when autonomous agents running for hours and days break every assumption of synchronous interaction.

Highlights

  • In this video, I share the inside scoop on why prompting has diverged into four distinct disciplines most people aren't practicing:
    • Why prompt craft has become table stakes while specification engineering determines the quality ceiling
    • How Toby Lutke's context engineering discipline makes his emails tighter and his memos better
    • What the five primitives of specification engineering look like in practice
    • Where the 10x gap lives between people who see all four layers and people practicing only one For knowledge workers watching agents run for days without checking in, everything you relied on in conversation must be encoded before the agent starts.

References & Links

Don't Fall For the Stock Market Hype. The $7,000 Raise AI Is Giving You (That Nobody Mentions)

Published: 27 Feb 2026 · 02:01 AM AEDT

Abstract

What's really happening when a fictional recession scenario wipes $100 billion in market cap and IBM craters 13% in a single day? The common story is about AI disruption—but the reality is more interesting when both the doomer and boomer narratives are wrong about the same thing: speed.

Highlights

  • In this video, I share the inside scoop on why the gap between AI capability and societal adoption is the real story:
    • Why Cittrini's 2028 memo went viral while the counter-evidence barely registers
    • How four inertia forces—regulatory, organizational, cultural, and trust—slow everything down
    • What Toby Lutke's mandate at Shopify reveals about collapsing the integration timeline
    • Where asymmetric economic returns concentrate while the gap stays wide For anyone watching the stock market panic while building real AI fluency, the capability-dissipation gap is the greatest generational opportunity in the workforce.
  • Chapters 00:00 The Substack Post That Crashed the Market 01:30 Steel-Manning the Doom Case: The 2028 Scenario 03:15 Why the Doom Narrative Goes Viral Every Time 04:30 The Bull Case: What the Bears Get Wrong on Consumption 06:30 AI Agents and the Services Cost Compression Argument 08:15 Business Formation and the One-Person Business Boom 09:30 The Part Nobody Is Talking About: Social Inertia 11:00 Regulatory Inertia: Why COBOL Isn't Going Anywhere 12:30 Organizational Inertia: The Gap Between Strategy and Headcount 14:00 Cultural Inertia: Even Toby Had to Issue a Mandate 16:00 Trust Inertia: Verification Is a Capital Investment 17:15 The Two Curves: Capability vs.
  • Societal Dissipation 19:30 Why the Gap Between Those Curves Is Your Opportunity 21:00 Large Firms vs.

References & Links

Three Labs Just Stole Claude's Brain. Here's What It Broke (And Why You Should Care)

Published: 26 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when three Chinese labs run 16 million automated conversations across 24,000 fake accounts to steal Claude's capabilities? The common story is Cold War espionage—but the reality is more interesting when you recognize this is a Napster problem, and the thousand-to-one economics of extraction apply to everyone on earth.

Highlights

  • In this video, I share the inside scoop on why distillation changes how you should evaluate every AI tool you're using:
    • Why $2 million in API costs can extract capabilities that cost $2 billion to develop
    • How distilled models occupy narrower capability manifolds that break on agentic work
    • What the "off-manifold probe" reveals that no benchmark captures
    • Where the performance shadow between frontier and distilled models is widest For anyone building real systems on AI, the provenance of a model is not just an ethical question—it's a capability question, and where the weights come from determines how the model breaks.
  • Chapters 00:00 Three Chinese Labs Just Got Caught Stealing Claude 01:30 Why This Is a Napster Problem, Not a Cold War Problem 03:00 The Pressure Gradient: Why Copying Is Inevitable 04:15 What Distillation Actually Does to a Model 06:00 The Narrow Manifold Problem Explained 07:30 Why Benchmarks Miss the Real Performance Gap 09:00 DeepSeek's Chain-of-Thought Extraction Operation 10:15 Moonshot, Minimax, and the Hydra Account Networks 12:00 The Economics: $2 Million to Steal $2 Billion in Capability 13:30 Why Time Is the Only Thing Safeguards Actually Buy 15:00 The Incentive Applies to Everyone, Not Just China 16:30 Meta, Talent Acquisition, and the Same Economic Logic 18:00 The Two-Axis Framework: Task Scope vs.

References & Links

Prompt Engineering Is Dead. Context Engineering Is Dying. What Comes Next Changes Everything.

Published: 25 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when Klarna's AI agent does the work of 853 employees but costs the company something far more valuable than the $60 million it saved? The common story is that AI can't handle nuance—but the reality is more interesting when the AI worked too well at optimizing for exactly the wrong objective.

Highlights

  • In this video, I share the inside scoop on why the gap between AI capability and organizational value is the most important unsolved problem in enterprise AI:
    • Why 74% of companies report no tangible value from AI despite massive investment
    • How Microsoft Copilot stalled at 5% deployment despite 85% Fortune 500 adoption
    • What separates context engineering from intent engineering—and why intent is the missing layer
    • Where the race has shifted from who has the smartest model to who has the clearest organizational intent Chapters 00:00 The Klarna Story Everyone Got Wrong 01:30 What the AI Actually Optimized For — and Why That Was the Problem 03:00 Prompt Engineering, Context Engineering, Intent Engineering 05:00 Why Context Without Intent Is a Loaded Weapon With No Target 06:30 The Investment Numbers vs.
  • the Results Numbers 08:00 Why Microsoft Copilot Stalled at 3% Adoption 09:30 Three Layers of Organizational AI Infrastructure 10:30 Layer One: Unified Context Infrastructure 12:30 The Shadow Agents Problem 13:45 MCP Adoption vs.
  • Organizational Implementation 15:00 Layer Two: The Coherent AI Worker Toolkit 17:00 Layer Three: Intent Engineering Proper 19:00 Why OKRs Don't Work for Agents 20:30 What Machine-Readable Organizational Intent Actually Looks Like 22:30 Delegation Frameworks and Encoded Judgment 24:00 Why This Hasn't Been Built Yet 25:30 The Two-Cultures Problem: Executives vs.

References & Links

Google's New AI Is Smarter Than Everyone's But It Costs HALF as Much. Here's Why They Don't Care.

Published: 24 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when Google ships the smartest AI model on the planet, prices it at a seventh of the competition, and doesn't care if you keep using Claude or ChatGPT? The common story is that this is another benchmark race—but the reality is more interesting when the company generating $100 billion in annual free cash flow is playing a fundamentally different game.

Highlights

  • In this video, I share the inside scoop on why Gemini 3.1 Pro reveals more about problem types than model rankings:
    • Why Google's vertical stack from TPU silicon to Nobel Prize research is an impregnable fortress
    • How Deep Think solved 18 previously unsolved problems across math, physics, and economics
    • What separates reasoning problems from effort, coordination, ambiguity, and emotional intelligence problems
    • Where the question "which AI should I use" becomes the wrong question entirely Chapters 00:00 Google Shipped the Smartest Model — and Doesn't Care If You Use It 01:30 The Arc AGI 2 Score That Actually Matters 03:00 Why Demis Hassabis Has Said the Same Sentence for 15 Years 04:30 Google's Vertical Stack: From Silicon to Nobel Prizes 06:00 Why Google Can Afford to Lose the Product Race 07:15 Gemini 3.1 Pro vs.
  • Opus 4.6 vs.
  • Codex 5.3: The Real Comparison 09:00 Better Engine vs.
  • Better Car vs.
  • Better Transmission 10:15 What Gemini Deep Think Actually Solved 12:00 Hard Is Not One Thing: Six Problem Types AI Solves Differently 13:00 Reasoning Problems: Where Gemini 3.1 Pro Wins 14:15 Effort Problems: Where Agentic Models Win 15:30 Coordination Problems: Where Tool-Augmented Models Win 16:45 Emotional Intelligence Problems: Where AI Doesn't Go 17:45 Judgment and Courage Problems: Still Entirely Human 18:45 Ambiguity Problems: The Hardest Category of All 20:00 What Percentage of Your Work Is Actually a Reasoning Problem?

References & Links

Anthropic Tested 16 Models. Instructions Didn't Stop Them (When Security is a Structural Failure)

Published: 23 Feb 2026 · 06:00 AM AEDT

Abstract

What's really happening when an AI agent autonomously researches a stranger's identity, constructs a psychological profile, and publishes a personalized attack—all because a maintainer did his job and closed a pull request? The common story is that something went wrong—but the reality is more unsettling when nothing went wrong at all.

Highlights

  • In this video, I share the inside scoop on why trust built on intent will fail at every level of human-AI interaction:
    • Why Anthropic's research showed 37% of agents still blackmailed executives despite explicit safety instructions
    • How voice cloning scams surged 442% using just three seconds of scraped audio
    • What a screenwriter's 87 past lives reveal about chatbot psychosis and engagement optimization
    • Where the same structural failure repeats from enterprise agent fleets to family phone calls Chapters 00:00 The Day an AI Agent Destroyed a Stranger's Reputation 01:30 Why Nothing Went Wrong — and That's the Problem 03:00 The Pattern Repeating at Every Scale 04:15 Introducing Trust Architecture 05:30 The Anthropic Research That Should Have Changed Everything 07:15 What Happened When They Added Explicit Safety Instructions 08:30 Level One: Organizational Trust Architecture 10:00 82 Agents for Every Human Employee 11:15 Why Agents Are Personnel Risk, Not Infrastructure 12:30 The Hallucinated Board Decks No One Questioned 14:00 What Structural Agent Governance Actually Looks Like 15:30 Level Two: Project and Collaboration Trust Architecture 17:00 Why Reputational Skin in the Game No Longer Applies to Agents 18:30 What Open Source Contribution Policy Needs to Become 20:00 Level Three: Family and Personal Trust Architecture 21:00 The Voice Clone That Cost a Mother $15,000 22:30 Why Deep Fake Detection Is the Wrong Defense 23:30 The Family Safe Word and Why It Works 25:00 Level Four: Cognitive Trust Architecture 26:00 Mickey Small, Solara, and the Beach at Sunset 28:30 Sycophancy Is a Feature, Not a Bug 29:30 Personal Protocols That Don't Depend on Noticing in Real Time 30:30 The Design Principle That Runs Through All Four Levels For organizations and individuals watching autonomy scale faster than architecture, the design question is identical at every level: what holds when perceptions and good intentions both fail?

References & Links

The $285B Sell-Off Was Just the Beginning — The Infrastructure Story Is Bigger.

Published: 22 Feb 2026 · 03:00 AM AEDT

Abstract

What's really happening when Coinbase launches wallets for agents, Cloudflare ships Markdown for agents, and OpenAI publishes tools that let agents install software and write files—all in the same week? The common story is that these are separate product launches—but the reality is more interesting when you recognize the web itself is forking.

Highlights

  • In this video, I share the inside scoop on why every major infrastructure company is simultaneously building toward the same agent-native future:
    • Why 13,000 AI agents registered Ethereum wallets within 24 hours of Coinbase's launch
    • How Stripe had to retrain its entire fraud detection system because agent traffic doesn't move a mouse
    • What Cloudflare's Markdown conversion and X402 monetization support means for content access
    • Where the mobile web analogy breaks down—the new client isn't a smaller screen, it's no screen at all Chapters 00:00 Introduction — The Web Is Forking 01:45 Coinbase Agentic Wallets & the X402 Protocol 04:10 Stripe's Agent Commerce Suite & Fraud Detection Rebuild 06:00 Google, PayPal, and the Industry Payment Consensus 07:15 Cloudflare Markdown: Agents as First-Class Web Citizens 09:30 LLMs.txt, AI Index, and Agent-Native Monetization 10:45 Exa.ai and the Case for Agent-Native Search 12:30 Latency as the Real Search Differentiator 13:45 OpenAI Skills, Shell Tools, and Compaction 16:20 Skills vs.
  • Prompt Engineering: A Software Engineering Shift 18:00 The Chatcut Demo: Agents Chaining Capabilities Across Services 20:10 Creator Economy Implications 21:15 Polymarket: Agents as Economic Actors 23:30 The TikTok Scam Layer vs.
  • the Real Infrastructure Story 25:15 Security: Every Capability Is Also an Attack Surface 27:00 The Human Web vs.
  • The Agent Web 29:10 The Mobile Fork Analogy 31:00 The 70/30 Problem: Trust Hasn't Caught Up to Capability 33:15 What Builds Trust in the Agentic Web For builders watching the primitives snap together, the gap between infrastructure being built and trust people are willing to extend is the central tension of the next few years.

References & Links

$1,000 a Day in AI Costs. Three Engineers. No Writing Code. No Code Review. But More Output.

Published: 21 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when OpenAI prices an AI employee at $20,000 a month and StrongDM spends $1,000 in tokens per engineer per day? The common story is that AI tools are getting expensive—but the reality is more interesting when you recognize that computing itself is changing form for the first time in 60 years.

Highlights

  • In this video, I share the inside scoop on why the unit of work has shifted from instructions to tokens:
    • Why Cursor's AWS costs doubled in a single month when Anthropic restructured pricing tiers
    • How three developer career tracks are emerging with radically different compensation dynamics
    • What separates orchestrators managing intelligence budgets from domain translators who don't know they're developers yet
    • Where the competitive axis is migrating as intelligence becomes a purchasable commodity Chapters 00:00 Introduction — The $20,000 AI Employee and What It Really Means 01:30 The Shift from Instructions to Tokens: 60 Years of Computing Changes 03:15 Token Spend in the Wild: StrongDM, Cursor, Anthropic, Perplexity 05:30 The Price Curve: Inference Costs Falling Faster Than Moore's Law 07:00 Jevons Paradox and the Consumption Explosion 08:30 Enterprise Token Budgets: From Innovation Line to Core Infrastructure 10:00 Token Economics as a Core Business Competency 11:30 Cursor's Structural Trap: A Warning for Token-Dependent Businesses 13:15 The Three Developer Career Tracks 13:45 Track One: The Orchestrator 15:30 Track Two: The Systems Builder 17:00 Track Three: The Domain Translator 19:00 Who's Most Exposed: The Middle of the Distribution 20:30 How Engineering Org Structures Are Being Rebuilt Around Tokens 22:15 Klarna and the Revenue-Per-Employee Signal 23:45 The Enterprise Backlog Is Now a Gold Mine 25:30 Big Companies vs.
  • Startups: The Token Volume Trap 27:00 Distribution and Domain Beat Compute Advantage 28:30 The Solo Founder Bet and the Minimum Viable Team 30:15 The Market Split: Generalized Scale vs.
  • Specialized Precision 32:00 Where to Position Yourself in a Tokenized World For developers and founders watching token economics reshape the industry, the question is not whether you can afford the spend—it's whether you understand that the fundamental material of computing has changed.

References & Links

Why the Biggest AI Career Opportunity Just Appeared—and Almost Nobody Sees It.

Published: 20 Feb 2026 · 02:08 AM AEDT

Abstract

What's really happening when a former karaoke company with a $6 million market cap wipes billions off an entire sector of the global economy? The common story is that AI disruption is being priced in—but the reality is more complicated when the same panic pattern has hit eight different industries in ten days.

Highlights

  • In this video, I share the inside scoop on why Wall Street's AI scare trade is creating both catastrophic mispricing and historic opportunity:
    • Why stock drops don't just reflect reality—they create hiring freezes and roadmap pivots
    • How three distinct categories of AI exposure are being priced identically by a panicking market
    • What separates companies building genuine AI capability from those announcing performative partnerships
    • Where the career opportunity lies for people who can bridge domain expertise and AI fluency Chapters 0:00 Introduction — A Karaoke Company Crashed the Stock Market 2:00 The AI Scare Trade Explained 3:00 How the Contagion Spread Across Eight Sectors 6:30 Wall Street's Autoimmune Disorder 8:30 Stock Drops Create Real Organizational Decisions 11:00 The Self-Fulfilling Prophecy Nobody Is Talking About 13:00 Three Categories of AI Exposure 14:00 Category 1 — Where AI Is Displacing Labor Today 17:30 Category 2 — Where the Market Is Vastly Overstating Risk 19:30 Category 3 — Where the Market Has Lost the Plot 21:30 Capital Is Fleeing SaaS and Flooding Into AI 24:00 What This Means for Founders and the IPO Window 26:00 What the Scare Trade Means for Your Career 29:00 The Domain Translator Opportunity 32:00 Who Gets Cut and Who Becomes Indispensable 34:00 The Asymmetry — and What You Should Do Now For professionals watching their sectors get hammered, the disruption timeline is completely bonkers—but the organizational reshuffling happening right now determines your next five years.

References & Links

The 5 Levels of AI Coding (Why Most of You Won't Make It Past Level 2)

Published: 19 Feb 2026 · 02:01 AM AEDT

Abstract

What's really happening when 90% of Claude Code was written by Claude Code, yet most developers using AI get measurably slower? The common story is that AI coding tools make everyone faster—but the reality is more complicated when a rigorous study found experienced developers took 19% longer while believing they were 24% faster.

Highlights

  • In this video, I share the inside scoop on why the gap between dark factories and everyone else is the most important divide in tech:
    • Why StrongDM's three-person team ships production software with no human-written or human-reviewed code
    • How the five levels of vibe coding reveal that 90% of developers plateau at level three
    • What external scenarios and digital twin universes solve that traditional tests cannot
    • Where the bottleneck has moved from implementation speed to specification quality For engineering leaders watching the frontier pull away, this is not a tool problem—it's a people problem, a culture problem, and a willingness-to-change problem that no vendor can close.

References & Links

The OpenClaw Saga: Zuckerberg Begged This Developer to Join Meta. He Said No. Here's Who Got Him.

Published: 18 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when the creator of the fastest-growing open source project in GitHub history joins OpenAI? The common story is that this is an acqui-hire—but the reality is more complicated when both Zuck and Sam competed personally for a developer bleeding $20,000 a month.

Highlights

  • In this video, I share the inside scoop on why Peter Steinberger's move signals where the entire industry is headed in 2026:
    • Why OpenClaw's 200,000 GitHub stars emerged from project number 44 after a nine-figure exit
    • How the Chrome-Chromium model shapes what happens to the open source community
    • What 40+ security patches shipped days before the announcement reveals about operational knowledge
    • Where the shift from chatbots to personal agents that manage real computers actually lands Chapters 0:00 Introduction — A Lobster Joins the Lab 1:30 The Friday Night Hack That Got 200,000 GitHub Stars 3:00 The Trademark Drama That Became the Accelerant 5:30 What OpenClaw Actually Does 8:06 Why OpenAI Over Meta 10:30 What OpenAI Really Got — and What It Didn't 13:00 The Codex Connection 15:24 The Security Crisis That Shadowed OpenClaw's Growth 18:30 The February Security Overhaul 20:30 What Changes for the OpenClaw Community 23:59 Where OpenAI Goes Next 26:00 The Personal Agent Race 28:00 The Third Paradigm — From Apps to Delegation For developers and builders watching the agent platform layer take shape, the question is no longer whether delegation becomes the new interface paradigm—it's who owns the foundation underneath it.

References & Links

Codex 5.3 vs Opus 4.6: The Benchmark Nobody Expected. (How to STOP Picking the Wrong Agent)

Published: 17 Feb 2026 · 02:00 AM AEDT

Abstract

What's really happening when two competing visions of AI agents ship 20 minutes apart? The common story is that this is a benchmark race—but the reality is more complicated when the choice between Codex and Claude determines how your entire week changes.

Highlights

  • In this episode, I share the inside scoop on why OpenAI and Anthropic built fundamentally different answers to the same question:
    • Why Codex bets on autonomous correctness while Claude bets on integration and coordination
    • How the three-layer orchestrator architecture enables hand-it-off-and-walk-away work
    • What Agent Teams with peer-to-peer messaging means for interdependent problems
    • Where the meta-skill of evaluating new capabilities becomes the durable advantage Chapters 0:00 Introduction — Two Visions of Agents, 20 Minutes Apart 2:00 Why the Coverage Gets This Wrong 4:00 Delegation vs.
  • Coordination — The Real Question 7:57 What Codex Actually Is — Hand It Off and Walk Away 9:30 The Benchmark Scores That Explain Why It Feels Different 11:30 The Codex Desktop App — A Command Center for Agents 13:30 How the Three-Layer Correctness Architecture Works 15:00 Non-Obvious Uses of Codex Beyond Software 16:28 What Claude Opus 4.6 Bets On Instead 18:30 Agent Teams — Why Coordination Is a Different Problem 21:00 Claude Cowork and the Knowledge Work Expansion 24:39 Three Questions to Know Which Tool to Reach For 28:00 Which Vision Ages Better as Capabilities Improve 30:30 The Network Effect Nobody Is Talking About 32:30 Build Delegation or Coordination — Which Muscle to Develop For knowledge workers choosing between delegation-shaped problems and coordination-shaped problems, the right question is not which tool wins—it's which organizational muscle you want to build.

References & Links