Rework

Roughly ninety short chapters, most one or two pages, each stating a position and defending it briefly. There is no framework, no sequence and no attempt at completeness.

The authors' credibility rests on having run a deliberately small, profitable, bootstrapped software company for years — which makes the book a report from an alternative rather than a theory about one.

Not a method. A set of positions from people who did it the other way and stayed solvent.

Devmark’s reading — not a quotation from the book

Their most quoted position: plans are guesses, and calling a guess a plan gives it an authority it has not earned. Long-range planning, they argue, commits you to decisions made when you knew least.

The qualification readers miss is that they are not against thinking ahead; they are against treating a document as a commitment and letting it override what you learn. Their alternative is deciding what to do this week and doing it.

They treat headcount, funding and office space as costs rather than achievements, and argue that many businesses grow past their natural size because growth is what businesses are supposed to do.

Their position is that small is a viable permanent state, not a stage. This was genuinely unfashionable when written and has since become a recognised position, partly because of this book.

A recurring theme: limited time, money and people force prioritisation that abundance removes. Their advice to build half a product rather than a half-built product follows from it.

The related instruction — start with the epicentre, the thing without which the product does not exist — is the most practically useful idea in the book and applies well beyond software.

Build half a product, not a half-built one. Constraints do the prioritising that abundance prevents.

Devmark’s reading — not a quotation from the book

Their treatment of the working day is the part that has been most widely adopted: meetings are expensive because the cost is multiplied by attendance, interruption is the enemy of real work, and long uninterrupted stretches are the unit of productivity.

Their alternative practices — asynchronous written communication, a protected quiet period, fewer and shorter meetings with an agenda — are now reasonably common, which they were not in 2010.

They argue for underdoing the competition rather than matching feature lists, for saying no by default, and for building for yourself rather than for a hypothetical customer.

The build-for-yourself position is the most contestable. It works when the founders are genuinely in the market they serve and fails badly when they are not, and the book does not much acknowledge the second case.

Their hiring positions: hire late, hire managers of one who need no supervision, ignore résumés and test with real work, and treat a long tenure at a large company as a warning rather than a credential.

On culture they argue it cannot be installed by decree — it is the accumulated behaviour of the company over time, and a values document that does not describe what actually happens is worse than nothing.

The book is written from a specific position — a bootstrapped software company with high margins, no investors and no capital requirements — and much of the advice depends on that position being available.

A business with real capital needs, inventory, regulatory obligations or network effects that reward speed cannot simply decline to plan or decline to grow. The authors present their constraints as universal virtues, and they are neither.

It is written to provoke, which makes it readable and occasionally makes it sloppy. Several positions are stated in a strong form the authors clearly do not hold absolutely, and a reader taking them literally will do something foolish.

The most common failure mode is using "planning is guessing" to justify not thinking, which is not the argument being made.

The durable contributions are the working-practice ones: interruption is expensive, meetings cost attendance multiplied by duration, long stretches are what produce work, and saying no is a default rather than a rudeness.

And the epicentre question, which is the fastest way to find out what a product actually is. Read it in an hour, keep four or five positions, and be sceptical of the rest.