Demis Hassibis is most creative at 1AM

Google DeepMind CEO says he starts his second workday at 10 p.m. and com... https://t.co/FzwrCTAaYR via @YouTube

— Bill Benzon, BAM! Bootstrapping Artificial Minds (@bbenzon) April 19, 2026

Biphasic Sleep vs. a continuous 8 hours

While the soul of this message is true—we shouldn't pathologize every waking moment—the history is a bit more complex.

- Biphasic sleep was real. Pre-industrial humans did sleep in two shifts. We lost this not to a mattress conspiracy, but to artificial light, which compressed…

— Emeka Sepi Maduka (@Sepiyoung) April 5, 2026

Some anecdotal evidence for segmented sleep: When I've been up during the night in December 2024

Over the years I've developed an interest in segmented sleep, as it is called. Instead of sleeping straight through the night for eight hours, or seven or nine, whatever, you have several sleep sessions interspersed with sessions where you do other stuff. This happens to me during a creative phase. I went into a down phase a couple of months ago and started coming out in the third week of November. I've had a LOT of segmented sleep since then. 

One of the things I'd do in one of my middle-of-the-night work sessions is make an entry on my Facebook page indicating that I was awake. I don't know just when I started doing that, but the first time I did it I posted a video clip of one of those "Time to make the donuts" commercials from the 1980s. I've been doing that every once in awhile since then. But other things as well.

This time I decided to keep a systematic record. Here's the record for December, 2024. I was up during every night except the 6th, 10th, and 13th. In some nights I was up more3 than once. I'll try to keep a record through January and then we'll see.

Time	Date	Content of Facebook notice
4:18 AM	12.1.24	Awake in the middle of the night. Good thoughts. All’s well.
3:22 AM	12.2.24	Middle of the night. Awake. Good thoughts.
4:02 AM	12.3.24	4:02 AM Tue 12.3.24 [I simply noted the time]
4:30 AM	12.4.24	Just another marker.
3:36 AM	12.5.24	3:38 AM & I’m just cruising by.
5:20 AM	12.7.24	Still at it.
4:33 AM	12.8.24	It’s that time again. [plus a photo of a donut]
3:18 AM	12.9.24	How’s everybody doin’?
4:16 AM	12.11.24	Peek-A-Boo!
5:41 AM	12.12.24	on the trail of CONSCIOUSNESS
4:32 AM	12.14.24	Someone’s been busy [+a photograph of a box of Krispy Kremes]
12:45 AM	12.15.24	12:45 AM
3:00 AM	12.15.24	Imagine this is about donuts, that's my state of mind at 3:00AM. [Attached to a video of “Shrimp Boats.”]
5:01 AM	12.15.24	Would someone please pass the donuts.
4:34 AM	12.16.24	Time to check in.
3:32 AM	12.17.24	Just a quick one, I promise. Comment added: Umm....er... Not so quick. 3:32 to 4:43. & I make two blog posts!
4:02 AM	12.18.24	Me to Self: You just HAD to check our email?
2:28 AM	12.19.29	2:28 AM and the night is still young.
4:09 AM	12.19.29	Nag Nag Nag! U do the donuts, I got work to do. (4:09)
4:02 AM	12.20.24	Psssst. Anyone here? {4:02 AM}
4:49 AM	12.21.24	Up every night in Dec. so far but 6, 10 & 13.
12:46 AM	12.22.24	12:46 AM and the night is still young. [Note: I’d gone to bed at about 8:15.]
3:17 AM	12.22.24	It’s been 2 1/2 hours since last check in.
5:14 AM	12.22.24	#3
2:38 AM	12.23.24	12° at 2:40AM in Hoboken
4:33 AM	12.23.24	11° at 4:34AM in Hoboken
2:54 AM	12.24.24	I tried to resist, I held out as long as I could. 2:54 AM
		Comment to above: & now it's 3:51 AM. Back to bed...& maybe to sleep as well.
3:46 AM	12.25.24	Gotta put cookies out for the Big Guy. 3:46 AM
2:10 AM	12.26.24	Resistance is futile. 2:10 AM
4:22 AM	12.26.24	Curse those Borg! 4:22 AM
4:21 AM	12.27.24	What time is it, kids? [Howdy Doody opening]
4:09 AM	12.28.24	Not Doris Day. 4:09 AM [Clip of Pixies singing “Que Sera Sera.”]
2:46 AM	12.29.24	I cannot believe...
4:49 AM	12.29.24	AGAIN? 4:49AM
4:24 AM	12.30.24	Right on time: 4:24 AM
1:21 AM	12.31.24	Peek-a-boo!

Coming out of a melancholy period [down phase]

I’m want to continue investigating my blogging pattern as I’ve discussed it in my recent 3 Quarks Daily article, Melancholy and Growth: Toward a Mindcraft for an Emerging World. This time I want to my daily activity for October and November.

This chart shows my daily posting activity during October and November. It’s been sparse during October, with many days where I didn’t post at all.

I had 30 posts in October, 15 in September, 18 in August, and 61 in July. So, I started going down in July, hit bottom in August and September, and started back up in November. Starting on the 15th I’ve posted every day – and I’m continuing to post daily in December.

That’s not all. I’ve gotten relatively little sleep for the last few days. I’d wake up during the night, start thinking, and then go to the computer and look around, maybe even write some stuff and do a post or two. Then I’d go back to bed. Maybe I’d sleep, but maybe I’d just lie in bed and think, sometimes on my back, which is not generally how I sleep.

[Note, by the way, that the practice of sleeping straight through the night is a cultural convention, not a biological necessity or preference. It’s called “segmented sleep.” I’ve got a number of posts about it.]

There’s more. During some of those night sessions I’d make a post on Facebook indicating I was awake. This started on November 22, where I posted a Dunkin’ Donuts commercial from the 1980s. This is one of those commercials featuring an owner of a Dunkin’ Donuts who awakens in the early morning, uttering “Time to make the donuts.” Here’s a list of those posts: first the time, then the date, and then the content of the post.

3:37 AM 11.22.24 – [Post a Dunkin’ Donuts TV commercial]
2:07 AM 11.24.24 – Again, I'm waking up in the middle of the night and thinking. What's next? Writing?
3:24 AM 11.24.24 – I’ve thought. Now back to sleep.
4:35 AM 11.24.24 – Back at it. More thinking.
11:31 AM 11.24.24 – I’ve spent the last two hours immersed in a remarkable line of thought. More later.
11:37 PM 11.25.24 – It’s been a productive day.
4:18 AM 12.1.24 – Awake in the middle of the night. Good thoughts. All’s well.
3:22 AM 12.2.24 – Middle of the night. Awake. Good thoughts.
4:02 AM 12.3.24 – 4:02 AM Tue 12.3.24 [I simply noted the time]
4:30 AM 12.4.24 – Just another marker.

Note that I’ve highlighted two of them. The content of the posts is like that of the other posts, but the first one went up late in the morning while the second went up late at night, after my normal bed time.I believe that first post, on the 25th, is when I began thinking about the implications for splintered career paths as AI becomes more deeply embedded in society.

The thoughts and ideas I have been having during this period are new and different. In particular this is the period where I decided, first, to write a 3QD article about my blogging habits, and second, to actually write the article. That article is a substantially new kind of intellectual activity for me.

If, as I’ve argued in my 3QD piece, those down-times are for the kind of mental reorganization that precedes a creative phase, one would expect new things to emerge [tumble out] at the start of an up-cycle.

Has this happened at the beginning of every up-cycle back to 2010, when I started blogging? I don’t know. I’ve not been keeping those kinds of records. I could go through the blog and take note of the content of every post to determine whether or not new stuff shows up immediately after every down cycle. That would take time, but I’m not going to do it now. I’ve got other things to do.

But looking at my blog posts won’t tell me anything about changes in my daily routine of the kind I’ve reported above. But I have been through this kind of thing before. It’s not new to me. It will slow down – it already is – it always does. And there are other night-notices scattered in Facebook, going a few years back. But I’m not going to look for them now.

Things to do.

More later.

The "two-sleep" system (two periods of sleep at night, separated by wakefulness)

Zaria Gorvett, The forgotten medieval habit of 'two sleeps', BBC Future, Jan 9, 2022. Much of the article is apparently based on, Roger Ekrich, At Day's Close: A History of Nighttime.

In the 17th Century, a night of sleep went something like this.

From as early as 21:00 to 23:00, those fortunate enough to afford them would begin flopping onto mattresses stuffed with straw or rags – alternatively it might have contained feathers, if they were wealthy – ready to sleep for a couple of hours. (At the bottom of the social ladder, people would have to make do with nestling down on a scattering of heather or, worse, a bare earth floor – possibly even without a blanket.)

At the time, most people slept communally, and often found themselves snuggled up with a cosy assortment of bedbugs, fleas, lice, family members, friends, servants and – if they were travelling – total strangers.

To minimise any awkwardness, sleep involved a number of strict social conventions, such as avoiding physical contact or too much fidgeting, and there were designated sleeping positions. For example, female children would typically lie at one side of the bed, with the oldest nearest the wall, followed by the mother and father, then male children – again arranged by age – then non-family members.

A couple of hours later, people would begin rousing from this initial slumber. The night-time wakefulness usually lasted from around 23:00 to about 01:00, depending on what time they went to bed. It was not generally caused by noise or other disturbances in the night – and neither was it initiated by any kind of alarm (these were only invented in 1787, by an American man who – somewhat ironically – needed to wake up on time to sell clocks). Instead, the waking happened entirely naturally, just as it does in the morning.

The period of wakefulness that followed was known as "the watch" – and it was a surprisingly useful window in which to get things done. "[The records] describe how people did just about anything and everything after they awakened from their first sleep," says Ekirch.

Under the weak glow of the Moon, stars, and oil lamps or "rush lights" – a kind of candle for ordinary households, made from the waxed stems of rushes – people would tend to ordinary tasks, such as adding wood to the fire, taking remedies, or going to urinate (often into the fire itself).

For peasants, waking up meant getting back down to more serious work – whether this involved venturing out to check on farm animals or carrying out household chores, such as patching cloth, combing wool or peeling the rushes to be burned. One servant Ekirch came across even brewed a batch of beer for her Westmorland employer one night, between midnight and 02:00. Naturally, criminals took the opportunity to skulk around and make trouble – like the murderer in Yorkshire.

But the watch was also a time for religion.

For Christians, there were elaborate prayers to be completed, with specific ones prescribed for this exact parcel of time. One father called it the most "profitable" hour, when – after digesting your dinner and casting off the labours of the world – "no one will look for you except for God”.

Those of a philosophical disposition, meanwhile, might use the watch as a peaceful moment to ruminate on life and ponder new ideas. In the late 18th Century, a London tradesman even invented a special device for remembering all your most searing nightly insights – a "nocturnal remembrancer", which consisted of an enclosed pad of parchment with a horizontal opening that could be used as a writing guide.

But most of all, the watch was useful for socialising – and for sex.

As Ekirch explains in his book, At Day's Close: A History of Nighttime, people would often just stay in bed and chat. And during those strange twilight hours, bedfellows could share a level of informality and casual conversation that was hard to achieve during the day.

For husbands and wives who managed to navigate the logistics of sharing a bed with others, it was also a convenient interval for physical intimacy – if they'd had a long day of manual labour, the first sleep took the edge off their exhaustion and the period afterwards was thought to be an excellent time to conceive copious numbers of children.

Once people had been awake for a couple of hours, they'd usually head back to bed. This next step was considered a "morning" sleep and might last until dawn, or later. Just as today, when people finally woke up for good depended on what time they went to bed.

Ekirch also references bi-phasic sleep in the classical era.

Biphasic sleep is common among among animals as well.

"There are broad swaths of variability among primates, in terms of how they distribute their activity throughout the 24-hour period," says David Samson, director of the sleep and human evolution laboratory at the University of Toronto Mississauga, Canada. And if double-sleeping is natural for some lemurs, he wondered: might it be the way we evolved to sleep too?

The move away from biphasic sleep happened during the Industrial Revolution:

"Artificial illumination became more prevalent, and more powerful – first there was gas [lighting], which was introduced for the first time ever in London," says Ekirch, "and then, of course, electric lighting toward the end of the century. And in addition to altering people's circadian rhythms. artificial illumination also naturally allowed people to stay up later."

However, though people weren't going to bed at 21:00 anymore, they still had to wake up at the same time in the morning – so their rest was truncated. Ekirch believes that this made their sleep deeper, because it was compressed.

As well as altering the population's circadian rhythms, the artificial lighting lengthened the first sleep, and shortened the second. "And I was able to trace [this], almost decade by decade, over the course of the 19th Century," says Ekirch.

I mention multiphasic sleep in other posts, such as:

• Reading The Human Swarm 5: Night-time action and brain-to-brain coupling [Shazam!!]
• Sleep: 8 continuous hours is a cultural convention, not a natural need

H/t Tyler Cowen.

Reading The Human Swarm 5: Night-time action and brain-to-brain coupling [Shazam!!]

Ch12 of The Human Swarm, “Sensing Others” starts off with two dynamite paragraphs. I say they’re dynamite because they speak to two of my hobby horses, a relatively recent one, and one that goes back to my graduate school days. Let’s dig in (p. 161):

Spend time among hunter-gatherers and you realize sleeping through the night is a modern idea. On my trip to Namibia, I listened as Bushmen conversed under a chiseled Milky Way in voices elaborate as birds, full of clicks, twang, and twitter. Their huts were barely visible in the flicker of family fires as they shared both traditional tales and evens of the day with gusto and moments of dramatic acting. When the sun was up their conversations centered on daily business. Nighttime was story time, with tales conveying the big picture of a proper social life and affirming people’s connections with the greater society.

That’s my recent hobby horse, and it is grounded in my own behavior. While I’m basically an “early to bed, early to rise” guy, in recent years I’ve noticed that in creative phases I may get up once or twice in the night and work (creatively) for an hour or two. Some posts:

• Sleep, NOT, an 8-hour block (9.27.12)
• Sleeping Habits (mine – 11.15.12)
• Sleep Ain’t What It used to Be (1.12.13)
• Sleep: 8 continuous hours is a cultural convention, not a natural need (5.14.14)

Moffett’s second paragraph is about brain-to-brain communication:

Years later their animated storytelling came vividly to mind as I stood behind Uri Hasson, a professor of psychology and neuroscience at Princeton University, while he hunched over a computer displaying a series of brain scans. Hassonhad been observing the cerebral activity of people watching a movie. The snippets he played were open to interpretation: some viewers might suspect the husband in the film was unfaithful while others might conclude the wife was a liar. When Hasson looked at scans of the viewers’ brains, he found they differed accordingly. If, however, audience members talked to each other as they watched a video, their cerebral cortexes synchronized: the same portions of their brains lit up when they began to follow the story line with a unified point of view. Hasson calls this mind-to-mind coupling “a mechanism for creating and sharing a social world.” The pleasure the Bushmen expressed that starry night must have come from such a melding of minds.

Moffett sites an article I’d posted here at New Savanna in 2013, Brain-to-brain coupling: a mechanism for creating and sharing a social world. I’ve posted a number of articles by Hasson. More generally, I’ve got a good many posts on the topics of coupling and synchrony (you’ll find there’s a fair amount of overlap between those two topics). There is now a large and growing literature on the neural activity of people interacting with one another.

My interest in this topic goes back to my graduate school days, when David Hays got me interested in the work of William Condon, who’d taken high-speed video of people intereacting with one another and observed that their movments where synchronized on a scale of 10s of milliseconds. This then became the central conception of my 2001 book on music, Beethoven’s Anvil, where I devoted the second and third chapters to developing the dual notions that, 1) tighly coupled synchronzed movement is central to music-making, and 2) and music-making is central to human society. I developed these ideas in the context of Walter Freeman’s conception of the nervous system as a complex dynamical system.

Now I’ve got to get back to reading Moffett’s book. More later.

More on segmented sleep

My friend Barry Solow just alerted me to this, Letter of Recommendation: Segmented Sleep, in the NYTimes:

Back when segmented sleep was common, this period between “first” and “second” sleep inspired reverence. The French called it dorveille, or wakesleep, a hypnotic state. English speakers called it “the watch.” I had usually approached the post-­midnight hours full-sail, by staying up. Waking into them is different, childlike. The time feels freer. The urge to be busy abates. Conversation has a conspiratorial intimacy, as if you’ve sneaked behind the tent to find the only other smoker at the wedding. Though I preferred the name dorveille, because it sounded glamorous, “the watch” was technically more accurate during those early weeks, when I mainly got up and watched Netflix.

In the preindustrial West, most people slept in two discrete blocks and used dorveille for all kinds of purposes. Having sex was popular. Benjamin Franklin liked to “take cold-air baths,” a fancy way of saying “open his windows naked.” Many people wrote in journals or interpreted dreams, which feel more proximate at 3 a.m. than in daylight. You could drink a cup of tea and take a satisfied leak on the embers of your fire. But everyone did something, because dorveille was ubiquitous.

We tend to view our basic patterns, like spending the evening out and then lying unconscious for eight hours straight, as belonging to a natural order. But until streetlights, being up late meant wandering town in butt-­clenching terror, tripping over stray animals until the wind blew out your lantern and you were set upon by armed bandits. So you went home before dark and went to bed early, waking after midnight.

After all, we're not born to sleep thru the night. Teaching/coaxing/cajoling  infants to sleep through the night is one of the major tasks of childhood. As I recall from something I read years ago, we're born to a 3 hour cycle, 90 minutes awake, and the 90s minutes of sleep. That's what has to be shaped to a "normal" sleeping pattern.

More on sleep at New Savanna. & sleep is just one behavioral mode among many.

Sleep: 8 continuous hours is a cultural convention, not a natural need

T. M. Luhrmann in the NYTimes:

This obsession with eight hours of continuous sleep is largely a creation of the electrified age. Back when night fell for, on average, half of each 24 hours, people slept in phases. In “At Day’s Close,” a remarkable history of night in the early modern West, Roger Ekirch writes that people fell asleep not long after dark for the “first sleep.” Then they awoke, somnolent but not asleep, often around midnight, when for a few hours they talked, read, prayed, had sex, brewed beer or burgled. Then they went back to sleep for a shorter period. Mr. Ekirch concludes, “There is every reason to believe that segmented sleep, such as many wild animals exhibit, had long been the natural pattern of our slumber before the modern age, with a provenance as old as humankind.”

A different regime, for example:

What have we lost with our dismissal of what the writer George Sturt called the “quiet depths of darkness”? In traditional non-Western societies like the Toraja, what happens at night really matters. People pay close attention to their dreams, and because they are awakened more often, they have more opportunity to remember them. When the anthropologist Eduardo Kohn arrived in a small village deep in the Amazon, people slept largely outdoors in an open thatch house, surrounded by other people. They would wake at night to drink tea, because it was cold, or because of the calls of animals. “Thanks to these continuous disruptions,” he writes, “dreams spill into wakefulness and wakefulness into dreams in a way that entangles them both.”

Sleep Ain't What It Used to Be

Back in September I posted a paragraph from a NYTimes op-ed arguing that the 8-hour block IS NOT a 'natural' sleep pattern. It's culturally imposed. There's more where that came from:

* * * * *

Matthew J. Wolf-Meyer. The Slumbering Masses: Sleep, Medicine, and Modern American Life. University of Minnesota Press, 2012.

Jacket copy [boldface mine]:

Americans spend billions of dollars every year on drugs, therapy, and other remedies trying to get a good night’s sleep. Anxieties about not getting enough sleep and the impact of sleeplessness on productivity, health, and happiness pervade medical opinion, the workplace, and popular culture. In The Slumbering Masses, Matthew J. Wolf-Meyer addresses the phenomenon of sleep and sleeplessness in the United States, tracing the influence of medicine and industrial capitalism on the sleeping habits of Americans from the nineteenth century to the present. 

Before the introduction of factory shift work, Americans enjoyed a range of sleeping practices, most commonly two nightly periods of rest supplemented by daytime naps. The new sleeping regimen—eight uninterrupted hours of sleep at night—led to the pathologization of other ways of sleeping. Arguing that the current model of sleep is rooted not in biology but in industrial capitalism’s relentless need for productivity, The Slumbering Masses examines so-called Z-drugs that promote sleep, the use of both legal and illicit stimulants to combat sleepiness, and the contemporary politics of time. Wolf-Meyer concludes by exploring the extremes of sleep, from cases of perpetual sleeplessness and the sleepwalking defense in criminal courts to military experiments with ultra-short periods of sleep. 

Drawing on untapped archival sources and long-term ethnographic research with people who both experience and treat sleep abnormalities, Wolf-Meyer analyzes and sharply critiques how sleep and its supposed disorders are understood and treated. By recognizing the variety and limits of sleep, he contends, we can establish more flexible expectations about sleep and, ultimately, subvert the damage of sleep pathology and industrial control on our lives.

Sleeping Habits

I've been adjusting my sleeping habits.

Like most of us, I grew up believing that one must get eight hours of sleep a night and that those hours had to be in a continuous chunk. The latter was rarely if ever explicitly stated; it was just assumed. Eight hours is eight hours straight.

So, when I found myself nodding off at 8, 8:30, or 9 PM I'd set myself some task that would keep me awake at least until 9:30 or 10. And when I'd waken in the middle of the night I just lay there in bed until I went back to sleep. Sometimes that happened quickly and sometimes it didn't.

And I'd take frequent naps during the day.

Then I read David Randall's article in the NYTimes. Shazaam! You mean it doesn't have to be eight continuous hours? You mean this 8-hours thing is a recent invention, that people routinely awaken in the middle of night and do stuff? That there was a lot of napping in the world, I already knew that, and did it myself. But the other stuff? Hot-diggity-dawg!

Now when I feel sleepy at 8PM I go to bed. And when I awaken at midnight or 1PM or 2PM or whenever, I'll get up and work for an hour or two. No problems.

And the napping continues.

Sleep, NOT an 8-hour block

David Randall has an interesting op-ed on sleep in the NYTimes:

Typically, mention of our ever increasing sleeplessness is followed by calls for earlier bedtimes and a longer night’s sleep. But this directive may be part of the problem. Rather than helping us to get more rest, the tyranny of the eight-hour block reinforces a narrow conception of sleep and how we should approach it. Some of the time we spend tossing and turning may even result from misconceptions about sleep and our bodily needs: in fact neither our bodies nor our brains are built for the roughly one-third of our lives that we spend in bed.

The idea that we should sleep in eight-hour chunks is relatively recent. The world’s population sleeps in various and surprising ways. Millions of Chinese workers continue to put their heads on their desks for a nap of an hour or so after lunch, for example, and daytime napping is common from India to Spain.