The 996 schedule means working from nine in the morning to nine at night, six days a week. It began as shorthand for a punishing work culture in Chinese technology companies and has since acquired admirers in San Francisco, where every old mistake eventually returns wearing a new vest.[1]

The argument for 996 is simple. Startups are races. The window is small. More hours create more output. The team willing to sacrifice more will outrun the team protecting its evenings.

Sometimes this is true.

A launch week can demand absurd hours. A production failure does not care about your evening routine. There are periods when the work is alive, everyone is pulling in the same direction, and stopping at six would feel less humane than continuing. I love those periods. I am not writing a defense of laziness. I am suspicious of any philosophy that makes exhaustion proof of seriousness, but I am equally suspicious of one that treats every difficult week as a labor violation.

The mistake is treating all work as the same kind of work.

Some work is additive. Ten customer calls usually teach you more than two. Fifty carefully written lines of code usually do more than five. More sales attempts create more chances to sell. There is a range in which more time really does buy more output.

Other work is nonlinear. Choosing the right problem, seeing the missing feature, finding the sentence that makes the product legible, noticing that the entire roadmap rests on a false assumption: these do not arrive in units per hour.

They often arrive after the hour.

Psychologists call one part of this incubation. You work on a problem, get stuck, leave it, and later return with a better answer or find that the answer has arrived while you were doing something else. A meta-analysis found a real incubation effect, especially for open-ended problems. Another study found that a simple task which allowed the mind to wander improved later creative performance.[2]

There is nothing mystical about this. A mind can remain attached to a problem without staring directly at it. It can loosen a bad representation, combine distant material, or notice an option the conscious search kept walking past.

The important word is can. A break does not guarantee brilliance. Most showers contain shampoo, not strategy. But removing every empty interval guarantees something else: the problem gets only the version of your mind that already failed to solve it.

I think of this as spare capacity.

Any system run at full capacity becomes fragile. When every hour is allocated, surprise becomes backlog. When every thought has a meeting attached, new information has nowhere to land. When the calendar is a wall of rectangles, the day can process planned work but cannot metabolize reality.

Founders often understand this in software and ignore it in themselves. They would never run critical infrastructure with no room for spikes, recovery, maintenance, or failure. Then they schedule the human system at one hundred percent and call the resulting latency commitment.

There is another cost. Switching tasks leaves what Sophie Leroy calls attention residue. Part of the mind remains stuck to the unfinished task, reducing performance on the next one.[3] A twelve-hour day split across Slack, recruiting, product, fundraising, and six minor emergencies may look heroic from the outside. Internally, it can be five half-minds passing a single keyboard between them.

Long hours also stop producing linearly. John Pencavel's work on factory output found that beyond a threshold, extra hours produced smaller gains.[4] Factory output is not product design, and a paper about wartime munitions should not be used as a magic number for software teams. The narrower point survives: hours and output are different variables. Treating them as identical is an accounting convenience, not a law of nature.

The mythology of 996 persists because hours are visible.

Judgment is not. Taste is not. Incubation is almost offensively invisible. A founder staring out of a window may be doing the highest-value work of the week or absolutely nothing. A founder sending messages at 1:14 a.m. is legible. The timestamp becomes a receipt for devotion.

Presence is easier to manage than thought, so organizations manage presence.

This is how a schedule built to produce commitment begins producing theater. People learn to remain online, write longer updates, create work that proves they are working, and delay leaving because leaving is the only act everyone can measure. The company gains hours and loses honesty.

The alternative is not a four-hour workday with a pottery class at noon. It is matching the schedule to the shape of the work.

Execution needs urgency. Discovery needs room. Craft needs long, unbroken attention. Recovery needs actual absence. A good week can contain all four. A bad culture flattens them into one demand: be here.

I am writing this partly against my own instincts. I like speed. I like unreasonable goals. I enjoy the particular joy of being tired in pursuit of my own choosing. I have also confused motion with progress enough times to know they are excellent impersonators.

The hardest question in a company is rarely "How can we do more?" The answer is always available. Add people. Add hours. Add meetings. Add features. The harder question is "What deserves to exist?"

More hours do not answer it. In some cases they delay the answer by making the current plan too expensive to question. Once a team has suffered for an idea, the suffering becomes evidence for the idea. Nobody wants to discover that the seventy-second hour was spent polishing the wrong thing.

This is why the best builders often look slower at the beginning. They are paying thought upfront. They are removing features before features acquire owners. They are waiting long enough to notice that the problem beneath the problem is the real one. Then, once the direction is right, they move with a violence that looks sudden only to people who missed the thinking.

There is a schedule that produces presence and a schedule that produces design. They may overlap. They are not the same.

By all means, sprint. Stay late for the launch. Work the weekend when the work has earned the weekend. There is a kind of collective intensity that is rare and beautiful, and I would never replace it with a policy document.

But intensity is a mode, not an identity.

The company should know when it is sprinting, why it is sprinting, and what recovery follows. Otherwise emergency becomes culture, culture becomes selection, and the people best at looking busy inherit the place from the people best at seeing.

You can manufacture effort. You cannot timesheet the leap.

A thousand phones on 996, perhaps.

Not one iPhone.