The fleet is already at sea.
Nobody has a map of it.
A quality engineer writes a tool on a Sunday, pays for it herself, tells no one, and quietly catches defects weeks earlier than her plant otherwise would. Not to cheat her employer — to protect a fragile idea from the machinery that would kill it in review. AI just made that machinery cheap enough for almost everyone to route around.
Why it hides
The best small ideas don't die because they're bad. They die because of whose patch they land on.
Radical ideas get killed by a fear of the new. But most trapped value isn't radical — it's small, sensible and useful. It dies for a quieter reason: it crosses into territory someone already owns, and gets defended against without ever being called a bad idea.
Why fixes fail
Every remedy has been tried for decades, and keeps quietly dying.
Sanctioned innovation time, suggestion schemes, skunkworks, labs. The solutions are old and well known. This book asks the question nobody does: if they work, why has none of them stuck?
Why hunting backfires
The obvious response — go and find the hidden work — is the most expensive mistake a leader can make.
Every instinct says surface it, govern it, bring it into the light. The evidence says the opposite. Hunt the fleet and it doesn't come up; it goes deeper, off company machines entirely. There is a way to see it. It isn't the one you'd expect.
“The only way to see the fleet is to stop hunting it.”
Every claim is traced to the research. The argument connecting them is the author's own, and labelled as such throughout. Every source is published, including the ones that cut against it → The register takes you to the papers themselves →
A book can't be corrected after it's published. Everything here can.
Every source, including the ones against me
Around a hundred studies, sorted by what kind of evidence each one is, with the disputed ones marked — plus the ten strongest attacks I could mount on my own argument. Four are still unanswered.
Read the research → The mechanismsThree moves, with the dials exposed
Chapter 12 puts three numbers in your hands: six weeks, five hundred euros, ten working days. They are starting values, not findings. This is where they live so they can move.
See the mechanisms → The registerWhere the research actually is
For every source: where to read the paper, what it found, and the passages in the book that draw on it. The studies doing the most work here are free to read in full, and the register takes you straight to them.
Open the register → The instrumentsThirteen exercises, one per chapter
Each chapter ends with something to run on your own organisation. Most take under an hour. Everything computes in your browser and nothing is sent anywhere.
Open the instruments →Thirteen things to run on your own organisation
Diagnosis is only half of it. Each chapter closes with a comparison, an audit or a piece of arithmetic that turns the argument on the building you actually work in. Some need an hour and a colleague in finance willing to be slightly indiscreet. None asks anyone to confess to anything, and that is deliberate.
One boundary holds across all of them: none is a tool for locating people running unsanctioned work. Every one points at your own practice. The book spends a chapter on why the hunt backfires — building the hunt here would be a strange way to end it.
Leapfrog
A corporate engineer's field guide to delivering AI on the cloud. Free, with nine runnable labs.
Submerged
Why organisations bury their own best ideas — and how to let the good ones surface without killing them.
