Bill Gates’ Think Week, Scaled Down to a Weekend

Try to remember the last stretch of hours you spent on where the business should be in three years, with nothing open and nobody able to reach you. If you cannot, this piece is for you. The Bill Gates Think Week was the extreme form of that stretch: seven days, twice a year, alone in a cottage on the water, reading papers Microsoft staff had written for him. The version that fits an operator’s life is two days, run on the same rules, with a smaller stack.

Most of what has been written about it treats the cabin as the trick. The cabin was the least important part.

Below is what the week consisted of, from the one reporter who was let in, then the weekend version we would run ourselves, and the two situations in which it is a waste of a Saturday and a Sunday.

What the Bill Gates Think Week Actually Was

In March 2005 the Wall Street Journal’s Robert A. Guth was allowed to visit, and his account, reprinted by the Seattle Times, is still the only detailed first-hand description.

The place was a secluded two-storey clapboard cottage on a quiet waterfront somewhere in the Pacific Northwest, reached by helicopter or seaplane. Gates had a small bedroom and an upstairs study facing the water, with a small fridge stocked with Diet Orange Crush and Diet Coke. Everyone was barred, family and staff included, apart from a caretaker who brought two simple meals a day. He read in bed, in the study, and at the kitchen table, which looked out at the Olympic Mountains.

The reading was the point, and the reading was not his own choice of material. Most of the papers were written by Microsoft engineers, executives and product managers through an open call, so anyone at any level could submit. About two months before the February week, his technical assistant Alex Gounares gathered submissions from across the company and set priorities. The database for that week held nearly 300 papers, with software security the largest category at 31. He aimed to get through at least 100 and his record was 112. Days ran to 18 hours, and some ran to 24.

The output was comments. After reading, he emailed his responses to the authors and sometimes to dozens of staff around the world, then wrote a summary for the executive team with a reading list attached. Guth quotes him on why that mattered: “If I write a comment that says, ‘We should do this,’ things will be re-orged, engineers will move.”

The best-known result came out of a 1995 week. “The Internet Tidal Wave” memo, dated 26 May 1995 and nine pages long according to a later Seattle Times account, turned the company toward the browser and the fight with Netscape. The National Constitution Center holds an excerpt with the line that did the work, “Now I assign the Internet the highest level of importance.” Guth also credits Think Weeks with helping along the Tablet PC, secure software and the online games business.

The habit began in the 1980s as quiet time at his grandmother’s house, reading and planning. By 2011 it had stopped being a solo exercise. Gates wrote on his own site that “we expanded who participates in Think Week a few years ago at Microsoft to include the top 50 engineering thinkers.”

What Made It Work, and What the Copies Miss

Strip the seaplane from the Bill Gates Think Week and three things remain, each one a design decision an operator can make cheaply.

The input was other people’s thinking. Gates went to the cottage to read the ideas his staff already had and decide which ones deserved his weight, which is a different exercise from a founder staring at a lake waiting for inspiration. It is closer to the method Charlie Munger describes, where you collect the models and arguments first and the judgement comes from having them arranged in one place. Our piece on Charlie Munger’s mental models covers that habit in more detail.

The output had authority behind it. A comment from Gates re-organised teams. A founder of a 20-person company can copy that part faithfully, because you have the same authority over your own firm. The week fails when the person doing it comes back with a mood and no decisions.

The intake was company-wide and open. The call for papers went to everyone, which meant the week doubled as the one channel through which a junior engineer could get an argument in front of the chief executive. Most founders who copy the week skip the intake and bring only their own material, which is why their version produces a tidier to-do list and nothing else.

One thing is better left alone. Eighteen-hour days of reading, some at 24, are a feat of stamina. Our piece on how to work smarter rather than harder makes the general case, and the specific one is that the Bill Gates Think Week delivered because of what was read and what was decided. The hours were a cost Gates could pay. Do not confuse the cost with the method.

The Weekend Version, Line by Line

The left column is what Guth saw. The right column is what we would run for a founder with a team of ten to fifty who can clear a Saturday and a Sunday twice a year and no more.

Element	Gates’ version (2005)	The weekend version
Length and frequency	Seven days, twice a year, February and a second week later in the year	Two full days, twice a year, booked six months ahead so nothing is scheduled over them
Place	A clapboard cottage on the water, reached by helicopter or seaplane	Anywhere that is not your home or office, within two hours’ travel, with a table, a window and poor reasons to leave
Company	Nobody, apart from a caretaker bringing two meals a day	Nobody. Meals arranged in advance so you never go looking for them
Material	Nearly 300 staff papers, gathered two months ahead by a technical assistant	Eight to twelve short submissions from your own people, requested four weeks ahead, plus your own five questions
Reading load	Target of 100 papers, record 112, days of 18 hours	All of the submissions once, slowly, and no more than two days of eight working hours each
Output	Emailed comments to authors and dozens of staff, plus a summary and reading list for executives	One memo to yourself of under 1,000 words, and three decisions sent to the people who have to act on them before Monday lunch
Who sees the result	The whole company, by design	Every person who submitted gets a reply, even a short one, within the week

What to Bring

Bring the submissions from your team. Four weeks before the weekend, ask every person who reports to you, and anyone else who wants to, for one page on a single question: what should we do differently in the next twelve months that we are not doing now, and why. One page, plain prose, no slides. You will get complaints, pitches and one or two things you had not seen, and all three are useful.

Bring your own questions, written in advance and limited to five, each specific enough to be answered. Which of our offers is weakest, and would we drop it if a competitor launched it tomorrow? Where does my own time go that someone else should own? Which customer would hurt most to lose, and what have we done for them this quarter? If you need a method for the first of those, run the weakest offer through the four variables in our piece on the Hormozi value equation before you go, and bring the result.

Bring the numbers on one printout, nothing live. Revenue by line, cash, pipeline, headcount, churn, a week old at most and frozen, because the weekend is for deciding what the numbers mean rather than watching them move.

Do not bring the shelf of titles on being more effective. Gates read what his engineers wrote about hard drives and language support, which is to say he read about his own business. Our round-up of productivity books is for a different kind of evening. One outside text is enough, chosen because it argues with something you believe, and read last.

What to Ban

Ban anything live. Email, messaging, the dashboards, the group chat. The test is whether the thing can deliver a new fact about this week. If it can, it stays behind, because every new fact pulls you back into operating and away from deciding.

This is a design problem rather than a discipline problem. Leaving the device in the car boot works where resolving to ignore it does not, which is the whole argument of our piece on how to increase willpower, which is to arrange the environment so the choice never comes up.

Ban calls, including from your deputy. Gates barred his family and his staff while running a company many times larger than yours. If two days of silence would endanger the firm, that is information, and the last section deals with it.

Ban the urge to produce. The strategy document can wait for the office, and the writing at the end is short by rule.

Where to Do It

Somewhere other than home, which has its own queue, and other than the office. A plain hotel, a borrowed cottage, a friend’s empty flat in another town, within two hours of the door so the travel is not the event. A table, a chair, a window onto something that moves slowly, and meals arranged so that food never becomes a reason to go out and find a signal.

Gates had the Olympic Mountains at the kitchen table. What did the work was that nobody could walk in and the reading was all there was to do. Choose for those two conditions first and the view second.

What Should Exist on Sunday Evening

Two things.

The first is a memo to yourself of under a thousand words. What you read, what surprised you, what you now believe about the next twelve months that you did not believe on Friday, and what you decided. Date it. You will read it again at the start of the next weekend, and the gap between what you believed and what happened is a more honest performance review than any you will be given.

The second is three decisions, sent to the people who have to act on them before Monday lunch. A weekend that ends with fourteen initiatives has ended with none. Warren Buffett’s method of ranking twenty-five goals and striking out all but five applies here with a lower cap, and our piece on the Warren Buffett 5/25 rule explains why the struck-out list is the dangerous one.

Each decision should name who owns it, what changes on Monday, and when you will check. If every item on your list is a question to investigate, you have deferred rather than decided, and the framework in our piece on decision-making in business is the place to go before the next attempt.

Then the replies. Every person who sent a page gets an answer within the week, even two lines. Your version earns a second round of submissions, the ones with the real ideas in them, only if the first round was seen to be read.

One caution on the memo. The temptation is to write it as the founding document of a new era. Famous routines get consumed as content, and our piece on Elon Musk’s daily routine is honest about how little of that transfers. The memo to yourself is private, so it has no reason to perform.

When a Think Weekend Is a Waste of Two Days

Two situations.

The company is in a cash crisis. If there are fewer than four months of runway and no agreed path to more, the next twelve months are not the question. The next six weeks are. A weekend of long-horizon thinking in that state is a comfortable way of avoiding the uncomfortable calls that need making on Monday morning. Make them, stabilise the cash, and book the weekend for the quarter after. Gates ran his weeks from a position where Microsoft could afford to be wrong about a trend for a year. You may not be able to afford a quarter.

You have not delegated enough to leave. If two days of silence would stall a sale, a payroll run or a delivery, the firm still depends on you for operating decisions that someone else should hold. That is a more urgent finding than anything you would read at the lake. Spend the weekend at your desk instead, working through who should own what, with our piece on how to delegate open beside you, and take the first proper weekend when the handover has held for a month. If nobody calls, you were ready.

A third case. If what you need is rest, take rest. A think weekend is two full days of concentrated reading and deciding, and it is no substitute for the thing described in our piece on how to take a day off. Using it as one will spoil both.

The person it suits is a founder past the survival stage, with a team that can run a fortnight without them, who has noticed that the three-year questions keep getting answered by default. Someone making a career change at 50, with a plan and no team yet to submit papers, can run a thinner version with their own five questions and a handful of letters from people whose judgement they trust.

Naval Ravikant’s observation that judgement is the scarce input, and that it compounds when you give it time, is the strongest case for doing this at all. His argument on how to get rich treats clear thinking as the one input that cannot be hired, and two days twice a year is the cheapest way we know to give it room.

Frequently Asked Questions

How long was the Bill Gates Think Week, and how often did he do it?

Seven days, twice a year, according to the Wall Street Journal’s 2005 visit, with one week in February. He spent them alone in a cottage on the water in the Pacific Northwest, reading papers written by Microsoft staff and sending comments back to the authors and the executive team.

What came out of the Think Weeks?

The most famous output was “The Internet Tidal Wave” memo of 26 May 1995, which assigned the Internet the highest level of importance at Microsoft and led to the browser push against Netscape. The 2005 account also credits the weeks with helping the Tablet PC, the company’s work on secure software and its online games business.

Can a founder with a small team realistically copy it?

Yes, at two days rather than seven, if the rules are kept: material from your own people requested weeks ahead, no live information, nobody able to reach you, and a short memo plus three decisions at the end. It is not worth doing during a cash crisis or before the business can run two days without you.

The cottage and the seaplane are the details people remember. What produced the memo was a founder reading what his own people thought, then telling them what he had decided, and that fits in two days.