Articles

Notes on transformation,
specification, and the immune system.

Direct, opinionated writing for CEOs navigating AI transformation. Published here in full — the canonical versions.

From HAL to Jarvis From HAL to Jarvis, Part 1: The Dreamers Wrote the Spec Leonardo drew the helicopter, Verne wrote the Moon shot, Asimov wrote the rules and Turing set the test. Each got the machine right and the hard part wrong. 8 October 2026 · 7 min read From HAL to Jarvis From HAL to Jarvis, Part 2: The Obedient Machine The golem, the genie, the sorcerer's broom and HAL 9000 share one fear. It was never the machine. It was the machine that does exactly what it is told. 8 October 2026 · 3 min read From HAL to Jarvis From HAL to Jarvis, Part 3: The Rule-Writers For thirty years, every rule in an AI system was typed by a human. It worked, and then it drowned in its own arithmetic. Why the first era of AI ended, and where it still runs. 8 October 2026 · 6 min read From HAL to Jarvis From HAL to Jarvis, Part 4: A Machine with Millions of Dials A neural network is a box with millions of dials that set themselves from examples. The idea dates from 1958, was buried twice, and needed two things the 1980s did not have. 8 October 2026 · 6 min read From HAL to Jarvis From HAL to Jarvis, Part 5: What a Language Model Is A language model has read everything and looks nothing up. What that means, how one is built, what it is good for, and the first rule of using one: put the source on the table. 8 October 2026 · 6 min read From HAL to Jarvis From HAL to Jarvis, Part 6: The Apprentice's Problem At home the assistant reads your files and works for hours. At the office the same model sits behind a narrower door. What careful means, and where the constraint has moved, as of October 2026. 8 October 2026 · 7 min read From HAL to Jarvis From HAL to Jarvis, Part 7: The Assistant Who Knows You The machine talks, and it is starting to act. What is missing is memory, permission and a written description of how you work, and none of the three is a model. 8 October 2026 · 5 min read Awake · Field The Audit Said It Would Not Work. What Proved It Did? A Field Awake on a platform I built and ran inside a Fortune 500 company. An external audit said the new serverless architecture was unproven and would never hold its cost estimate. It ran beside the old one for a year. Here is what the proof looked like, and what the people asking for proof up front looked like. 28 September 2026 · 21 min read AI transformation Vibe Engineering: You're Not Prompting AI. You're Building Your Business Harness. The discipline that makes AI coding agents reliable turns out to be business engineering: processes, gates, and evals for how your company actually works. 1 September 2026 · 11 min read Awake · Outside-in Did Volkswagen Build Its Future — or Another Supplier? An outside-in Awake of Volkswagen. The Rivian joint venture looks like a real edge — it owns the architecture, and it is earning production work. Whether it holds the right to inherit the work will decide what Volkswagen just built. 31 August 2026 · 23 min read AI transformation Is Your Company Ready for an AI Attack? The deepfake CFO, the perfect fake invoice, the instruction hidden in a PDF your own agent reads — they all travel through parts nobody owns. The same method that makes AI reliable makes you hard to attack. 31 August 2026 · 9 min read AI transformation The Smallest Business Harness One person delegating work to AI needs the same three institutions a company does: a constitution, a gates map, and a loop. The Edge Method at desk scale. 28 August 2026 · 8 min read AI transformation Delete Before You Automate Most AI projects automate a process that should not exist. The five-step algorithm, applied to a company: find every part nobody owns, cut it, and only then let the machine in. 27 August 2026 · 12 min read AI transformation The Corporate Immune System Is Killing Your AI Transformation The biggest threat to your AI strategy isn't the technology. It's the 50 reasons your organization will find not to change. 11 August 2026 · 9 min read AI transformation The Foundation Gap: Why Your AI Agents Don't Know What Your Company Values The most dangerous AI failure isn't disobedience — it's perfect execution of the wrong intent. 4 August 2026 · 8 min read AI transformation From Prompt Templates to Playbook OS: The Scaling Problem Nobody Talks About Structured prompts are a genuine step forward. They're also a ceiling you'll hit faster than you think. 28 July 2026 · 7 min read AI transformation Dan Shapiro's 5 Levels of Vibe Coding (and the Level 6 That Hits Every Job) Dan Shapiro's five levels run from spicy autocomplete to the dark factory — and most teams calling themselves AI-native are at Level 2. The honest map, plus Level 6: the same shift hitting every job in your company. 21 July 2026 · 8 min read AI transformation 20x Companies: How 5 Engineers Beat Teams of 100 The leanest companies in the world aren't cutting costs. They're multiplying what each person can do. 14 July 2026 · 7 min read AI transformation The J-Curve: Why AI Is Making Your Team Slower (And Why That's Temporary) Experienced developers are 19% slower with AI tools — while believing they're 24% faster. This gap explains everything about where we are right now. 7 July 2026 · 6 min read AI transformation The Specification Bottleneck: Why Your AI Strategy Is Failing Amazon built a tool that slows developers down. That tells you everything about where value is moving. 30 June 2026 · 7 min read AI transformation What Intelligence Analysts Know About AI That Executives Don't The most dangerous failure in AI isn't the machine that disobeys. It's the machine that executes a bad specification flawlessly. 23 June 2026 · 6 min read

Why the pilot worked and the company didn't

Field notes on stalled AI transformation, executive judgment, and the specification economy — the weekly letter behind Book 2. About once a week. No spam. Unsubscribe anytime.