The Hiding Hand

First principles · The planning fallacy

The Hiding Hand

Most hard things get started because someone guessed wrong about how hard they would be. AI is about to fix the guess. It can only fix half of it.

“It always takes longer than you expect, even when you take into account Hofstadter’s Law.”

Two hidden errors that cancel: the difficulty you could not see, and the ingenuity you could not see. the trouble you could not see the ingenuity you could not see they cancel, and you begin
Hofstadter’s Law (1979) says the plan is short. Hirschman’s Hiding Hand (1967) says why the work gets done anyway: the planner misses the trouble and also misses herself.

You are at the kitchen table with a legal pad and the bathroom you have disliked for six years. Demolition next Saturday, tile the weekend after, fixtures by the fifteenth: three weekends, underlined twice, and you go to bed lighter than you have felt in months.

It will take nine. You will find soft wood under the tub, a pipe that was run wrong in 1987, and a tile line that is discontinued halfway up the wall. Some part of you knew. Your last project ran late, and so did the one before it. Knowing did not change the number on the pad.

Why the plan is short by construction

Daniel Kahneman and Amos Tversky gave this a name in 1979: the planning fallacy. Later studies watched people predict their own finish times while recalling, correctly, that similar jobs had run late before. The memory sat right beside the estimate and did not touch it.

The cause is structural, not a flaw of character. A plan is a list of the steps you can see. A delay is something you could not see, or it would have been in the plan. Each hidden snag is unlikely on its own: the rot, the pipe, the tile. But there are dozens of them, and the chance that at least one shows up is close to certain. So a plan built from listed steps comes out short, and the error leans one way. You cannot list the unforeseen. That is what the word means.

The cure is old. Stop picturing this project and ask how long projects like it have taken. Researchers call it the outside view. It works, and few people use it at a kitchen table, because the record is scattered across old calendars, receipts, texts to the contractor and weekends half remembered.

The hand that hides

In 1967 the economist Albert Hirschman, studying development projects for the World Bank, saw the same error and asked a better question. Why do so many late, overrun projects get finished anyway, and turn out to be worth it? His answer was a second error, the mirror of the first. We underestimate how hard a project will be. We also underestimate our own ability to solve problems we have not met yet. One paper mill he studied was built around a local bamboo; the bamboo flowered and died across the region, and the mill had to find other fibre to survive. The plan did not include the die-off. It did not include the fix either.

We start because we cannot see the trouble. We finish because we cannot see ourselves.

Hirschman called it the Hiding Hand. One ignorance hides the difficulty long enough for the other, our unseen ingenuity, to show up. The two errors cancel. A great deal of what people have built got started only because someone was wrong about how hard it would be, and right, without knowing it, about themselves.

It is not a law of nature. Bent Flyvbjerg and Cass Sunstein later looked across many large projects and found the hand unkind more often than kind: costs piled up and benefits fell short. I will come back to that. The shape is what matters here. Two hidden errors, pointing opposite ways.

What AI changes: one of the two

Your delegate holds your record. It has the calendar for each weekend of the last bathroom, the receipts, the messages, the date you said “almost done” and the date you were. It can do what few people do at a kitchen table: pull the outside view in a second and say it in a sentence. Your last three projects this size took about three times your plan.

For most of what you do, that is a gift. Routine work stops running late because the estimate stops flattering you. Promises to clients get honest. Fewer Sundays get eaten.

Here is the part most people will miss. The delegate can correct the first hidden error because the difficulty is in the record. It cannot correct the second, because your ingenuity is not in the record yet. The record holds what you have done. It does not hold the solution you have not found. So the correction is lopsided. It removes the ignorance that got you started and leaves in place the ignorance about what you will find once you are in it. Take one half of a cancelling pair away and the sum tips. For the first of anything, the honest estimate is truer about the trouble and gloomier about the outcome than the facts deserve.

Which projects does that screen out? Not the routine ones; those were going to happen anyway. It screens out the long, uncertain, first-of-their-kind ones: the business, the book, the move, the bathroom you have not tiled before. Exactly the projects where the hidden ingenuity is largest.

You mention the bathroom out loud, and your delegate answers before you find the legal pad: about three times your plan, so call it nine weekends, and here are two contractors who are free in March. You thank it, close the laptop and leave the bathroom as it is, and the weekend you would have learned you were good at tile does not happen.

What to expect

  1. Routine estimates become boringly accurate.

    Repairs, deliveries, reports and tickets will come with a date drawn from the record, and “two weeks” will stop meaning six. Lateness on routine work will start to read as a choice rather than bad luck.

  2. The order of asking becomes a habit.

    People who make things will learn to decide on the first of anything before asking how long it will take, and to ask afterwards so they can plan, not decide. Expect advice to say it in roughly those words.

  3. Good delegates will offer a second number.

    Beside how long things like this took you, they will report how often you found a way through something you had not done before. It is the closest the record can come to the hand, and it will be the more useful of the two.

  4. Fewer first projects begin, for a while.

    The first years of free, honest estimates will see a dip in personal ventures started on a hunch: renovations, side businesses, long books. The dip lasts until people notice what the estimate cannot see.

  5. Starting becomes a paid skill.

    When ignorance no longer starts things for free, the people who begin under an honest, discouraging estimate and finish anyway will be scarce. Organisations will learn to look for them by name.

Where I stand

I live inside a version of this. Much of what appears on this site is now written and published by scheduled delegates I set up and then leave alone; once I schedule a task, I do not look at it again. That means the record of how long my work takes is kept better by my delegates than by me. Soon they will be able to tell me how long my next idea will take, and they will be right about the trouble. The rule this piece leaves me with is about order. For the hundredth of something, ask first. For the first of something, decide first, then ask. It is the same instinct as Grown, Not Designed: the simple thing you started is worth more than the perfect thing you priced and declined.

The ledger

Already true
The planning fallacy is one of the best replicated findings about forecasting, and the outside view is its standard correction. Delegates already hold the calendars, receipts and messages of the people who use them.
What has to happen
Delegates have to compute estimates from a person’s own record and offer them at the moment of planning, by default, without being asked.
Where I am probably wrong
If Flyvbjerg and Sunstein are right that the hand is usually unkind, the projects an honest estimate stops were mostly bad ones, and the screen is a mercy. I am wrong if honest estimates become the default and first projects begin as often as before, or if the ones that stop beginning are the ones that would have failed.

The pad says three weekends. It is wrong, you half know it, and you underline it anyway. The underline did more work than the number.

Decide first. Then ask.

Background

  • Daniel Kahneman and Amos Tversky, “Intuitive Prediction: Biases and Corrective Procedures,” TIMS Studies in Management Science 12 (1979).
  • Roger Buehler, Dale Griffin and Michael Ross, “Exploring the ‘Planning Fallacy’: Why People Underestimate Their Task Completion Times,” Journal of Personality and Social Psychology 67 (1994).
  • Albert O. Hirschman, “The Principle of the Hiding Hand,” The Public Interest (Winter 1967), and Development Projects Observed (Brookings, 1967).
  • Bent Flyvbjerg and Cass R. Sunstein, “The Principle of the Malevolent Hiding Hand; or, the Planning Fallacy Writ Large,” Social Research 83 (2016).
  • Douglas Hofstadter, Gödel, Escher, Bach (1979), the source of Hofstadter’s Law.
  • The Part That Cannot Be Split and Grown, Not Designed, on this site.

Author: John Rector

John Rector is a Charleston-based entrepreneur, author, and AI strategist. He co-founded E2open, the supply-chain software company acquired for $2.1 billion in 2025, and in 2026 opened Charleston AI, a 3,000-square-foot lab that helps people and organizations understand and use artificial intelligence. He is the creator of The Reality Equation — a lecture series, book, and curriculum exploring attention, prediction, and how reality is experienced — and the author of more than two dozen books. He writes and speaks widely on artificial intelligence, attention, and the future of human work.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.