The Founder Mode Delusion

The founder came to our team's stand-up on a Thursday in March, and for eleven days afterwards we were the most important team in the company.

It was genuinely good. He asked sharp questions, he knew things about our area that our own director didn't, and he unstuck a decision in four minutes that had been sitting for five weeks. I want to be completely straight about that, because what follows is not a story about a bad executive. He was excellent, in the room, on the day.

Then he didn't come back.

Not out of neglect — he went to another team, and presumably was excellent there too. But we had reorganised around eleven days of his attention. We had a decision he'd made in four minutes without the context that had made it take five weeks. We had a director whose authority had been quietly demoted to provisional, because everyone had now watched it be overridden. And we had no way to ask a follow-up question, because the mechanism that had reached us wasn't a process. It was a Thursday.

What the argument actually claims¶

Graham's Founder Mode, from September 2024, is a real argument and deserves stating properly rather than caricatured. Founders, he says, are handed a playbook written for professional managers — "hire good people and give them room to do their jobs" — and in Brian Chesky's case, "He followed this advice and the results were disastrous."

Manager mode, in Graham's description, is a specific structure: "you treat subtrees of the org chart as black boxes. You tell your direct reports what to do, and it's up to them to figure out how."

And the alternative has a structural consequence he names plainly:

“

it's going to break the principle that the CEO should engage with the company only via his or her direct reports

”

I think the diagnosis is largely right. Treating subtrees as black boxes does lose information, founders often do know things about the work that their org has filtered out, and the professional-manager costume genuinely fits some people badly.

The problem is what the essay treats as the binding constraint. Read it and the thing stopping founders is belief — bad advice, borrowed convention, insufficient nerve. Nowhere is the constraint a number.

The number¶

In 2006, Michael Porter and Nitin Nohria at Harvard Business School began tracking what CEOs actually do. Not surveys — observation, logged in 15-minute increments, 24 hours a day, seven days a week, eventually producing more than 60,000 hours of coded data across 27 CEOs.

Sixty-two and a half hours. Business conducted on 79% of weekend days and 70% of vacation days. Seventy-two percent of all working time already inside a meeting — and if you have read what meetings are actually for, you know most of that 72% is not time that could simply be reclaimed by being more efficient.

One-on-one42% of meetings

Small group (2–5 people)21% of meetings

Everything larger37% of meetings

Now put founder mode against that. Graham is asking a CEO to add a category of engagement — depth past the direct reports, into the subtrees. Where does the time come from?

There is no slack in a 62.5-hour week that is 72% meetings. Every hour spent three levels down is an hour removed from something else: a direct report, a customer, a decision, or the thinking time that is already the scarcest thing on the schedule.

What actually happens¶

Follow the arithmetic and the failure mode is predictable, which is why it looks the same at every company that tries this.

The founder picks a few areas. Not maliciously — by interest, by proximity, by whatever was on fire in March, by whichever demo they saw. There is no mechanism for choosing fairly because the choosing isn't a process, it's a calendar.

Those areas get intensity and whiplash. Real insight, real unblocking, and decisions made at a speed that is only possible because the person making them isn't carrying the context that made it slow. Some of those decisions are better than what the team would have reached. Some are the five-week problem being solved in four minutes by someone who doesn't know why it took five weeks.

Everyone else gets strictly less than before. This is the part nobody models. The un-visited majority don't stay where they were — they're worse off, because the layer above them has been publicly demoted. Their manager now runs a team where everyone has seen that any decision can be reversed by someone dropping in for an hour, and where nobody can predict whether this quarter is a visited quarter.

And the middle layer becomes provisional. Not removed, which would at least be legible — conditional. They keep the accountability and lose the finality. That's a specific and quite unpleasant job: you are answerable for outcomes you can be overruled on without notice, by someone whose engagement you cannot schedule. And this lands on a population that, as the research on untrained managers shows, mostly arrived in the role with no preparation for the ordinary version of it, let alone the provisional one.

That's a structure the formal chart never shows, and it's a good demonstration of why the published org chart was never the real one. Founder mode doesn't flatten a hierarchy. It adds a second, invisible hierarchy — proximity to the founder's attention — layered on top of the first, and the second one moves.

The turn¶

I've been arguing arithmetic, and arithmetic is easy to nod along to and hard to feel. So here is what it costs, concretely.

The one thing an employee needs from a management structure is to know whose answer is final. Not a good manager, ideally, but at minimum a knowable one. That single fact is what lets you commit to a plan, spend three weeks on something, tell a customer a date.

Founder mode, applied at scale to a 62.5-hour week, removes that fact without replacing it. Not because anyone intends to — because there are only so many hours, so the founder's engagement is necessarily sporadic, so finality becomes probabilistic. Your director's yes is a yes unless it isn't.

And the people who handle that best are the ones with the most slack: the ones who can afford to have three weeks of work reversed, who can wait out a bad quarter, who aren't counting how many days they could survive if this went wrong. Everyone else takes fewer risks, commits to less, and routes around the uncertainty — which produces exactly the cautious, low-ownership behaviour that founder mode was invoked to fix.

Nohria's own reported surprise from the study is worth sitting with here. As McKinsey summarised it, the single most common thing that shocked the CEOs was that they had assumed they spent lots of time with customers, and in fact few of them really did. These are people who were wrong about their own calendars. The proposal that they should now also reach three levels into the org, on top of a schedule they were already misjudging, is not a proposal about courage. It's a proposal about a resource nobody in the study had.

Working inside it¶

1. 1

   Find out whether you're in a visited area or not

   This is the first fact and almost nobody asks for it. Look at the last two quarters: how often did the founder engage directly with your team? Visited and un-visited are genuinely different jobs with different risks, and you can't plan until you know which one you have.

2. 2

   Get decisions in writing, with a name on them

   The specific damage of probabilistic finality is that verbal approval decays. "Confirming: we're doing X, per your call on Tuesday" costs one line and converts a decision into an artifact. It won't stop a reversal, but it makes the reversal visible as a reversal rather than as your misunderstanding.

3. 3

   If you're the middle layer, name the terms out loud

   The provisional position is survivable if it's acknowledged and corrosive if it isn't. "I'll own this, and if you're going to make calls in this area directly, I need to hear them within the day" is a reasonable ask. If the answer is no, you now know your job is coordination rather than decision — which is a real job, just not the one on your business card.

4. 4

   Batch what you'd want the founder to see

   Founder attention arrives sporadically and leaves quickly, so treat it as a scarce interrupt rather than a channel. Keep a standing short list of the decisions only they can unblock. When the Thursday comes, you spend it on the five-week problem instead of on a demo.

5. 5

   Watch what happens to the un-visited quarter

   The health of a founder-mode company is not visible in the areas the founder is in — those always look great. It's visible in the ones they haven't touched in six months. If those are drifting, the mode is being applied without the capacity, and the arithmetic above is the reason.

The Thursday¶

Our director left about seven months later, and in the exit conversation — the honest kind you only get once someone has nothing left to lose — she said the thing that has stuck with me longest about that period.

She said she never once minded being overruled. What she minded was not knowing, on any given Monday, whether she was running the team or holding it.

That's not a complaint about founder mode's ambition, and it isn't a defence of the black box either. It's a complaint about arithmetic being sold as attitude — about a 62.5-hour week being asked to cover a company, and everyone downstream absorbing the shortfall in the form of not knowing whose answer counts.

He was excellent, in the room, on the day. There were just never going to be enough days.

It's just business.

Sources

1. Paul Graham — 'Founder Mode' (September 2024)
   +How this was checked

   The essay: 'hire good people and give them room to do their jobs'; 'He followed this advice and the results were disastrous'; manager mode treating subtrees as black boxes; and that founder mode 'is going to break the principle that the CEO should engage with the company only via his or her direct reports'.

2. Michael E. Porter & Nitin Nohria — 'How CEOs Manage Time', Harvard Business Review, July–August 2018
   +How this was checked

   The study itself. HBR's page confirms directly that Porter and Nohria 'launched a study tracking how large companies' CEOs spent their time, 24/7, for 13 weeks' beginning in 2006; the article's detailed figures sit behind a paywall, so the numbers used here were taken from the two secondary reports below, which agree with each other.

3. MGMA — 'Test of time: Harvard study reveals how CEOs spend their days'
   +How this was checked

   Reports the study's figures: 27 CEOs, more than 60,000 hours coded, 15-minute increments 24/7; 9.7-hour average workday and 62.5-hour week; 37 meetings weekly at 72% of working time; one-on-ones 42% and small groups 21% of meetings; 46% of CEO time with internal employees spent with direct reports.

4. McKinsey — 'How do CEOs manage their time? The winners of the 60th annual HBR McKinsey Award explain'
   +How this was checked

   Corroborates 27 CEOs and 60,000 hours, and reports business conducted on 79% of weekend days and 70% of vacation days, plus Nohria's account that CEOs assumed they spent lots of time with customers when few actually did.