Food for Agile Thought #547: AI’s 1997 Internet Moment, Code Isn’t Product, Cognitive Surrender, Admitting Mistakes Is Not Enough

TL; DR: AI’s 1997 Internet Moment — Food for Agile Thought #547

Welcome to the 547th edition of the Food for Agile Thought newsletter, shared with 35,532 peers.

This week, Benedict Evans tells Lenny Rachitsky that today marks AI’s 1997 Internet Moment and asks whether automation kills tasks or jobs, which Casey Newton ’s guest, Kathryn Anne Edwards, treats as real but manageable, faulting unemployment insurance rather than fearing an idle underclass. Itamar Gilad warns that cheap AI coding tempts teams to build before validating, while Malcolm Spittler and Dylan Patel name the value GDP misses ‘Dark Output,’ and Joost Minnaar prefers autonomous-team networks over a single chain of command. Also, Marina Favaro and Jack Clark sketch the implications of AI's recursive self-improvement.

Next, Rich Mironov warns that AI ships 100x more code, yet attention and budgets don’t scale, so products that skip discovery rarely stick. METR, with Anthropic, Google, Meta, and OpenAI, finds that AI agents could plausibly go rogue but not robustly, while Teresa Torres notes that Cowork’s VM hosts the Mini Shai-Hulud worm without blocking it. Mike Fisher likens siloed teams to hand-copying the Diamond Sutra, and Jeff Gothelf redefines “done” as acceptable variance.

Lastly, Marc Abraham borrows venture capital’s ‘terminal value’ to help product managers judge whether a product merits more investment. Addy Osmani calls the opposite reflex ‘cognitive surrender,’ in which you stop thinking and accept the AI’s answer without checking it. Mark Graban shows confession works only when fixes follow, citing Burger King and Domino’s, while Tobi Lütke runs Shopify’s agent River in public Slack so everyone learns by watching. Finally, Anthropic open-sourced 11 role-specific Claude Cowork plugins for knowledge workers.

🎓 🇬🇧 The Claude Cowork Online Course — Available NEXT MONDAY (June 8) for $129

You have been prompting AI for months. The results are inconsistent, every conversation starts from zero, and the model forgets who you are. That is the ceiling of prompting.

The Claude Cowork Online Course teaches you to break through it: build Skills that encode your expertise, connect them to your tools, and assemble Agents who handle recurring work the way you would handle it yourself. No coding required.

What You Will Get:

✅ 8+ hours of self-paced video modules: Skills, Agents, delegation frameworks — ✅ Tested with a live BootCamp cohort (April 2026) — ✅ The A3 Framework: decide what to delegate and what to keep — ✅ Starter kit with folder structure, CLAUDE.md, and Skill templates — ✅ All texts, slides, prompts, graphics; you name it — ✅ Designed for the $20/month Pro plan — ✅ Lifetime access to the version you purchase — ✅ Claude Cowork Foundational Certificate.

👉 Please note: The course will be available for $129 from June 8 to 15, 2026! (After that, $199.) 👈

🎓 Join the Waitlist of the Course Now: Claude Cowork: Stop Prompting. Start Delegating. No Coding Required!

Did you miss the previous Food for Agile Thought issue 546?

🗞 Shall I notify you about articles like this one? Awesome! You can sign up here for the ‘Food for Agile Thought’ newsletter and join 35,000-plus subscribers.

🎓 Join Stefan in one of his upcoming training classes!

🏆 The Tip of the Week: AI’s 1997 Moment

📺 Lenny Rachitsky and Benedict Evans: A rational conversation on where AI is actually going

In conversation with Lenny Rachitsky, Benedict Evans believes AI is a ‘1997 internet moment,’ as big as the internet but only that big. He reframes job loss as ‘is this a task or a job?’, suggests distribution is becoming the real moat, and thinks things will probably turn out fine.

Source: 📺 A rational conversation on where AI is actually going

Authors: Lenny Rachitsky and Benedict Evans

🎯 Product

Rich Mironov: Code Isn’t Product

Rich Mironov warns that AI lets teams ship 100x more code, but customer attention, budgets, and buying don’t scale. Skip discovery, positioning, and marketing, and most new products won’t stick.

Source: Code Isn’t Product

Author: Rich Mironov

Itamar Gilad: Skateboards, Cars, and the Return of Code-First Discovery

Itamar Gilad warns that cheap AI coding is reviving the old habit of jumping straight to working prototypes, sidelining faster no-code validation. Building first still slows learning and breeds attachment to ideas.

Source: Skateboards, Cars, and the Return of Code-First Discovery

Author: Itamar Gilad

Marc Abraham: My Product Management Toolkit (70): Determining Terminal Value

Marc Abraham borrows ‘terminal value’ from venture capital to help product managers decide whether to keep investing in a product, combining payback period, a customer value scorecard, and strategic option value.

Source: My Product Management Toolkit (70): determining Terminal Value

Author: Marc Abraham

🧠 Artificial Intelligence

(via Anthropic): When AI builds itself

Marina Favaro and Jack Clark show that AI is already accelerating AI development inside Anthropic, where Claude now writes most of the code, then ask what recursive self-improvement might mean once machines start building their own successors.

Source: Anthropic: When AI builds itself

📺 Addy Osmani: ‘Cognitive surrender is when you stop thinking altogether and blindly accept the answer the AI gives you’

Addy Osmani names a failure mode he calls ‘cognitive surrender,’ the moment you stop thinking altogether and blindly accept whatever answer the AI hands you, rather than questioning or verifying it yourself.

Source: 📺 ‘Cognitive surrender is when you stop thinking altogether and blindly accept the answer the AI gives you’

Author: Addy Osmani

(via SemiAnalysis): AI Dark Output: The Visible Cost of Invisible Output

Malcolm Spittler and Dylan Patel propose that AI’s real output is becoming invisible to GDP, what they call ‘Dark Output.’ We can measure tokens spent and jobs lost, but not the value created.

Source: SemiAnalysis: AI Dark Output: The Visible Cost of Invisible Output

📖 (via METR): Frontier Risk Report (February to March 2026)

METR, with Anthropic, Google, Meta, and OpenAI, assessed whether internal AI agents could go rogue. The verdict: agents plausibly had the means, motive, and opportunity to start small rogue deployments, but couldn’t make them robust.

Source: METR: 📖 Frontier Risk Report (February to March 2026)

📺 Casey Newton and Kathryn Anne Edwards: An economist’s case against the AI jobs-pocalypse

Casey Newton interviews economist Kathryn Anne Edwards, who rejects the Silicon Valley ‘idle class’ fear as classist. AI job loss is real but absorbable; the fix is repairing unemployment insurance, not UBI.

Source: 📺 An economist’s case against the AI jobs-pocalypse

Authors: Casey Newton and Kathryn Anne Edwards

Teresa Torres: What I Learned from the Recent Wave of Package Hacks (And Is Cowork Immune?)

Teresa Torres examines the recent Mini Shai-Hulud worm and asks whether Cowork shields you. Its virtual machine contains malicious code, but does not block it, so guard what data you share.

Source: What I Learned from the Recent Wave of Package Hacks (And Is Cowork Immune?)

Author: Teresa Torres

🖥 💯 🇬🇧 Claude Cowork BootCamp #2, June 10-July 2, 2026

You bought the Claude Pro subscription. You installed the desktop app. You pointed Cowork at a folder, watched it churn for thirty seconds, and got something that looked impressive but was not quite useful. You tried again. Same result.

Most people who try Claude Cowork get stuck in the same place. They do not know which of their tasks are good candidates for automation. They do not know how to build Skills that survive a second use. They do not know where the current limits sit. So they keep treating Cowork like a chat tool and miss the point of having an AI assistant who actually does the work.

The Claude Cowork BootCamp fixes that. In four hands-on sessions, you build working Skills and AI Agents during the sessions, not after. You leave with a compounding system, not a stack of prompts.

The Founding Cohort sold out in April 2026. Cohort #2 starts June 10, 2026.

A Note on Positioning: This BootCamp is built for knowledge workers who want to automate repetitive work with Claude Cowork. There is no Agile-specific content in the curriculum. The class is in English. 🇬🇧

Learn more: 🖥 💯 🇬🇧 Claude Cowork BootCamp #2, June 10-July 2, 2026 — No Coding Required.

Customer Voice: “Vijay Reddy, Principal SPC & AI Governance Lead: “Three weeks ago I could use AI. Today I can deploy it. The Cowork Bootcamp is the only AI training I have taken that shifted my thinking from ‘what can I prompt?’ to ‘what should I architect?’ — and that shift showed up immediately: on the same day as Session 3, I shipped a production AI research agent live at sagent.sai4rai.org, applying Stefan’s CLAUDE.md principles and the A3 Framework in real code, not just in exercises. I would recommend this to any agile coach, product manager, or practitioner who is tired of AI training that teaches tools but leaves you without a system for knowing when and how to actually delegate to AI at scale.” (Vijay Reddy, Founder & Executive Director, SAI4RAI.)

➿ Agile & Leadership

Joost Minnaar (via Corporate Rebels): Chain of command: your organization’s biggest single point of failure

Joost Minnaar shows that a single chain of command is one point of failure. Drawing on Menger’s theorem, he proposes networks of autonomous teams whose many independent paths keep information flowing.

Source: Corporate Rebels: Chain of command: your organization’s biggest single point of failure

Author: Joost Minnaar

Mark Graban: Admitting Mistakes Is Not Enough: The Leadership Lesson Behind the Burger King and Domino’s Ads

Mark Graban shows why admitting mistakes alone falls flat: Burger King, Domino’s, and T-Mobile all paired public confessions with real fixes first, the same candor-plus-improvement sequence that builds psychological safety inside any team.

Source: Admitting Mistakes Is Not Enough: The Leadership Lesson Behind the Burger King and Domino’s Ads

Author: Mark Graban

Mike Fisher: From Scriptorium to System

Mike Fisher uses the ninth-century Diamond Sutra to argue that most companies are still hand-copying AI: siloed teams, drifting prompts, no shared standards. Leaders must decide when to stop experimenting and start standardizing.

Source: From Scriptorium to System

Author: Mike Fisher

📯 The Problem of AI Builders: Eric Ries’ ‘Incorruptible’ Solved the Later Problem. AI Builders Have an Earlier One.

Eric Ries’ new book ‘Incorruptible’ solves a problem most readers will not face for years: protecting a valuable organization from capture once it succeeds. The builders that AI is creating hit an earlier issue: Building software used to force the question of whether it was worth building. That gate has largely collapsed. Eric Ries asks how mission survives success; we, the normals, how judgment survives abundance.

Thesis: Ries’s Incorruptible solves a later problem, protecting a valuable organization from capture; cheap building created an earlier one, where judgment about what is worth building is the only gate left. That is the problem this article addresses.

Learn more: The Problem of AI Builders: Eric Ries’ ‘Incorruptible’ Solved the Later Problem. AI Builders Have an Earlier One.

🛠 Concepts, Practices, Tools & Measuring

Tobi Lütke: Learning on the Shop floor

Tobi Lütke explains why Shopify’s AI agent River only works in public Slack channels, turning the whole company into a Lehrwerkstatt where everyone learns by watching the most experienced people work alongside it.

Source: Learning on the Shop floor

Author: Tobi Lütke

Jeff Gothelf: What “done” means when you’re shipping AI features

Jeff Gothelf suggests that AI has broken our old definition of “done.” You are no longer shipping predictable vending machines, but distributions of unpredictable behavior, so done becomes a calibration of acceptable variance, not specification adherence.

Source: What “done” means when you’re shipping AI features

Author: Jeff Gothelf

(via Anthropic): Open source repository of plugins primarily intended for knowledge workers to use in Claude Cowork

Anthropic open-sourced 11 role-specific plugins for Claude Cowork, also compatible with Claude Code. Each bundle includes skills, connectors, slash commands, and sub-agents so that Claude can act as a specialist for functions such as sales, finance, legal, marketing, and data. Teams customize them with their own tools, terminology, and workflows for consistent results.

Source: Anthropic: Open source repository of plugins primarily intended for knowledge workers to use in Claude Cowork

📅 Training Classes, Meetups & Events 2026

Upcoming classes and events:

• 🖥 💯 🇬🇧 June 8–15 — Self-Paced Online Course: Claude Cowork: Stop Prompting. Start Delegating. (English; Self-paced Online Course)
• 🖥 💯 🇬🇧 July 1 — Meetup: HoA 75: Token Economics (English)
• 🖥 💯 🇬🇧 July 15— Self-Paced Online Course: AI 4 Agile Course v3 — Master AI for Agile Practitioners (English; Self-paced Online Course)
• 🖥 🇩🇪 August 25–16, 2026 — Live Virtual Class: Professional Scrum Product Owner Training (PSPO I; German)

👉 See all upcoming classes here

🗞️ The Previous Food for Agile Thought Edition

Read more: Food for Agile Thought 546: Choosing to Stay Human, Customer Research by LLM, AI Product-Market Fit, Enterprise Agility Today.

📺 Join 6,000-plus Agile Peers on YouTube

Now available on the Age-of-Product YouTube channel to improve learning, for example, about AI’s Labor Market Impact:

• Stop Writing Prompts. Let AI Do It for You — Hack #01, AI4Agile Online Course v2.
• Socratic Prompting — Hack #10, AI4Agile Online Course v2.
• Check Your AI’s Plan Before — Hack #7, AI4Agile Online Course v2.
• From Product Requirements to Experiments to Learnings — Supported by Generative AI.
• Never Accept an LLM’s First Offer — Improve GenAI’s Usefulness w/ Feedback Loops and Challenges.

🎓 Do You Want to Read More Like This?

Well, then:

• 📰 Join 35,000 peers and sign up for my weekly newsletter
• 📅 Join 6,000-plus peers of the Hands-on Agile Meetup group
• 🐦 Follow me on Twitter and subscribe to my blog, Age of Product
• 💬 Alternatively, join 20,000-plus peers of the Slack team “Hands-on Agile” for free.

Food for Agile Thought 547: AI’s 1997 Internet Moment, Code Isn’t Product, Cognitive Surrender, Admitting Mistakes Is Not Enough was first published on Age-of-Product.com.