An Algorithmic Solution to Insomnia (2017) | Hacker News Reader

An Algorithmic Solution to Insomnia (2017)

ilya.sukhar.com

109 pointsakkartik33 comments

I’ve had huge issues with falling asleep for years, and have had two big changes in the last 2 years that have completely changed things for me. The first was being diagnosed with anxiety and learning about mindful breathing activities, essentially just throwing on the headspace or calm apps while in bed allowed me to drift off very quickly after a 10-20min session.

The second big thing was having a baby. 7 months in and I can fall asleep anywhere at anytime simply because I’m so utterly exhausted all the time. I’d recommend trying the breathing first.

In my experience, anxiety-induced insomnia is an entirely different beast. I went through CBT-i and had a strongly negative experience: its strict rules made bed and bedtime feel like a prison (now I can't even lie quietly in my bed?!), and it made my insomnia even worse. More generally, "optimizing" for sleep, thinking about it at all, seems like a horribly counterproductive thing to do for somebody anxious.

Mindfulness meditation has been an enormous help for me also. It's not an exaggeration to say it's been life-changing. Vigorous exercise (it needs to be vigorous) as much as possible has also been great.

Can you recommend a path to start with mindfulness meditation?

Headspace (or similar apps? I haven't tried others) is great. If you're feeling up to it, the program I'd most strongly recommend is mindfulness-based stress reduction (MBSR), which asks for 45 minutes from you every day for 8 weeks. It was hard for me to be still for so long at first, but gets easier surprisingly quickly.

Thank you! Are you referring to this course or something else? https://palousemindfulness.com/

Get this book(let) pronto - https://www.penguinrandomhouse.com/books/545899/how-to-relax...

Did wonders for me in terms of helping me “empty my mind” and therefore relax, sleep well etc.

I would strongly recommend getting apps like Headspace, Calm after you’ve internalized the “theory” of the practice of relaxation.

Good luck!

What kills my sleeping habit is having different time for activities during the week and the weekend.

I sometimes lead a nice and regular workweek, but then comes the weekend and my social life throws my sleeping schedule all over the place. And suddenly I have insomnia again.

I tend not to have much sleepless nights, but when I do, I try to sleep like I was five years old. I remember really well how I used to go to sleep when I was a kid. I made pretend in bed. I pretended to be in a spaceship, or a fast car, or riding the train, and then I would seamlessly drift into sleep. Still works, but it is hard to make pretend, specially because an adults mind is not so easily entertained by itself.

What is one to do when you wake up at 3 am, all fresh, mind buzzing with ideas and problems ? You know you should sleep, otherwise your batteries run out of juice at noon, but you can’t. Is there an algo for that?

Listening to something you can focus on should help. I like meditation videos of at least one hour.

For me reading a dense non-fiction book will get me tired and back to sleep in 15-30 minutes. For my wife a cup of hot milk does it.

Wanted to do a top post, but since your situation is pretty much identical to mine...

I think the OP was wrong in avoiding pharmacological solutions out of hand. I 100% agree with his assertion that they come with side effects and addiction risks, but this only makes the game more complex. Throwing away a whole category of solutions is too wasteful.

First, CBT works. It worked for me as well, to a degree, although I suspect it's somewhat better when you have problems falling asleep than when you wake up too early (terminal or maintenance insomnia). Tip: most watches on the market (and my Oura ring) automatically calculate your sleep efficiency. So you don't need complex formulas - just limit your sleep every time sleep efficiency goes below a certain number (80%? varies from device to device). Another tip: at least for a while limit caffeine below what you think you should. Caffeine has an _average_ half life of 3 hours, but we're not all average - if in your case it's more like 6 hours, that by itself is enough to give you insomnia. And you'll never realize it because you only do what everybody else does. Third tip: relaxations exercises are rumored to work well, but only if you practice them for a couple of weeks.

Ok, pharmacological solutions. First, what worked for me. It turns out I have very weak anxiety, which, TBH, seems like a pretty common sense state to be in. The fix is to counter it with _very weak_ sedatives or anxiolytics, and to vary the dosage as needed, including day to day. For example, ~1mg of Mirtazapine (for reference, a pill is usually 30mg - I suggest grinding it with 30g of sugar and measuring grams). Side effects: slight afternoon somnolence and chocolate cravings if you take too much. Big advantage: no tolerance. In principle other mild sedatives would work as well, but I love its safety profile. It's an old school antihistamine, btw.

Ambien is ok to take when needed and I did take it for many years but... I don't like it. There are rumors of higher overall mortality, plus it hits your short term memory which makes it easy to overestimate how much it helped you sleep - you may just have forgotten you are awake. But it works, it's probably the safest benzo-like thing available and it goes out of your system in a few hours which makes it pretty good for morning insomnia. I still use it very occasionally. I found out that at least in my case it can be replaced with 0.5 mg of mirtazapine or Xanax. YMMV.

Melatonin had its own posts here. Short version: look for low dosage pills (1mg or less) and take them in two ways: either as a weaker but infinitely safer Ambien, or take it 3-5 hours before bed time to reset your circadian rhythm.

To stay awake: low dosage modafinil, aka 50mg or less. If you can feel the effects, you're taking too much. It's supposed to compensate for tiredness, not give you a high (which it will at 100mg and over). It goes well with a downer in the evening, like the 1mg of mirtazapine. Use your own wisdom on having a long term lifestyle of using pills to stay awake and fall asleep. In my case, I take it when I slept under 6 hours which tends to be about once a week, and it turns a bad work day into a very good day. I do not recommend it regularly or in higher dosages, just trust me on this. But it's awesome for the motivation: It helps a lot when I wake up groggy, check my Oura ring and instead of groaning I'm "yey, this is a low sleep day!".

CBT is supposed to replace drugs, but the version with very low dose modafinil and mirtazapine is a lot less painful.

Inositol is something I've just discovered and seems to click well with my brand of sleep issues. Of all the things I've tried it's absolutely the first which actually bumped the sleep efficiency over 90%. Good news: it's a supplement not a drug, and it's also supposed to be very safe. Bad news: dosage required for insomnia is pretty high, around 20g. Do your research - I've just started playing with it. But don't be surprised if lower dosages do nothing.

And a final word of advice: don't be afraid to experiment, within bounds. I'm testing solutions for almost 10 years now, but I've always had a current doctor to vet what I tried. My biggest input was playing with (usually much lower) dosages, and finding the mix which works for me.

I have an open notebook on my desk which I use through the day to jot down stuff to investigate, reminders about meetings and appointments etc. It's always open on the current day.

If I have this fresh idea in the middle of the night, my main anxiety seems to be that i'll forget it tomorrow, so I find just adding it to the list for tomorrow is enough to ease that concern, and allow me to get back to sleep.

Failing that, get up and get on with it, feel shit for the afternoon, but just keep going through to your normal bedtime, and obviously avoid power tools in the afternoon :)

I suffer from the same problem, if problem it can be called, it's just a different sleep pattern. I learned to live with it mostly. I wake up at 3 am and there is no way I fall asleep again before 1.5-2 hours. I just get up and do something productive, usually I'm quite creative at those hours and the silence is amazing.

The problem is that modern society isn't that nice with these uncommon sleep patterns, it's hugely problematic if you still need to wake up at something like 6 or 7 am.

I've read in the past that this is actually an ancestral pattern and that 7-8 hours of continuous sleep is something imposed to us with the industrial age.

I actually had no problem adapting my issue to modern society. It’s kids and dogs that make it impossible to sleep past 6 or 7am!

Wheezy Waiter has a video on trying biphasic sleep, and his conclusion was: If you wake up at 3, get up and do something. When you get tired, go back to sleep. Second sleep is the best sleep.

I tried that for a while and want to try it again, because it worked great for the week I tried it before falling out of the habit.

https://ilya.sukhar.com/blog/an-algorithmic-solution-to-inso...

I do something similar: often waking up at 3:30-4:30 AM. I just get up and start doing something that appeals to me without drinking coffee. About 2/3 of the time, I get a good 90 minutes of work or reading in and fall back to sleep for a few hours. 1/3 of the time, I just roll into my day and get tired slightly earlier that day.

On the days where I do sleep twice, 60% of the time I feel like Superman after the second sleep. (30% feels normal and 10% are brutal, usually where I had to wake to an alarm instead of naturally).

>Well, it turns out that insomnia is largely a form of performance anxiety that accrues over years of episodic poor sleep due to stress, travel, environmental changes, etc

While his scientific method is great, this guy is making the same mistake that most people make, which is thinking it is all in your head, and can be fixed by changing your thoughts. People make this mistake because they notice that when you are more relaxed, you get to sleep quicker and visa versa. This is true to a point. You can gain an extra 30 minutes or an hour if you are in the right mood, but this is only a small change either way of what your natural rhythm is already set to. You can't change it any more than that by thinking.

Sleep schedules are hard coded things that work by physical means. Long story short: wear full blue blocking glasses for 4 hours before bed at night (especially if you are using a computer), and a SAD light smack in your face first thing while eating breakfast in the morning, and you'll never have sleep trouble again. You can get very cheap versions of both on ebay or wherever, you don't need the fancy $100+ ones.

The gist of this seems be what is called "sleep compression" or sleep restriction the edges can be a bit blurry.

Sleep hygiene [1] is usually the first thing a doctor will advise.

If that does not work, or does not work well enough, your doctor will move on to sleep compression [3] or sleep restriction [2], or compression then restriction if compression does not work.

The next step is a "sleep study". You go to a lab, and they have a room you can sleep in, and they attach all sorts of sensors to your body. Getting up to pee is a big hassle for both you and the staff.

Based on the study results, they may try medication or a new round of sleep compression [3] or sleep restriction [2].

For me sleep compression [3] or sleep restriction [2] let to blinding headaches. My body requires more sleep so that method was not good for me.

There is no single procedure that will work for everyone. Sleep hygien has a high frequency of success. Sleep compression [3] or Sleep restriction [2] does to but less frequently. Listen to your body while you are experimenting and dont feel bad if the first step does not work.

I include a " Cautionary Note" from [2] sleep restriction therapy.

A Cautionary Note: It is suggested that you begin by spending a minimum of 5.5 hours in bed, regardless of how long you think you sleep, so that job performance does not suffer too much and you are able to function during the day. If you have significant difficulty staying awake, do not attempt this method. It is not for everyone. If you have any concerns or questions please talk to your doctor.

1. https://www.cdc.gov/sleep/about_sleep/sleep_hygiene.html

2. https://thrive.kaiserpermanente.org/care-near-you/northern-c...

3. https://www.med.upenn.edu/cbti/assets/user-content/documents... Keep sleeping less, until you do sleep.

I also found CBT-I useful but it doesn't solve the underlying problem (if there is one). For example, if Insomnia is a side effect of depression. Also, for leaving bed after 20 min if one doesn't fall a sleep. I got this advice as well - go read a book or something. But when I wake up at 4am the last thing I want to do is read a book. I prefer closing my eyes and get some rest / light sleep until my alarm goes off. While it's not sleep it's still restful. Listening to something with headphones helps me.

Discussed at the time:

Algorithmic Solution to My Insomnia - https://news.ycombinator.com/item?id=15997016 - Dec 2017 (79 comments)

I have the middle of the night insomnia that CBT-i doesn't seem to touch.

One minor thing I've done is to wake up to natural light (ie light or no shades or curtains). Prior, I had black out shades, and I'd sleep in but wake up groggy, then have trouble sleeping the next night. It made a pretty large difference for me (I wake up and sleep pretty early and feel a lot better). Obviously, if you need to stay up late this won't work for you.

There were other things I did that helped a ton (exercise, no caffeine, etc), but those are obvious and get mentioned a lot. I rarely see the natural light piece mentioned.

This is highly geographically dependent. For example, in Berlin where I live currently, sunrise in mid Summer is at 4:40am. I'm highly sensitive to light, and use blackout currently for this reason. That's not a remotely practical wakeup time with a regular work schedule.

Try vitamin D: https://pubmed.ncbi.nlm.nih.gov/28475473/

An important note is to take it in the morning, not at the evening. The article misses it somehow.

Intervention group received a 50 000-unit vitamin D supplement, one in a fortnight for 8 weeks

Not sure what this formula means, but 50,000 IU is way too much for a single dose. 10,000/day is prescribed only to eldery people who have enormous deficiencies on the edge of death, according to my endocrinologist. Check your D levels before eating any supplements.

https://www.nhs.uk/conditions/vitamins-and-minerals/vitamin-...

Do not take more than 100 micrograms (4,000 IU) of vitamin D a day as it could be harmful. This applies to adults, including pregnant and breastfeeding women and the elderly, and children aged 11 to 17 years.

Not true. I am prescribed 50,000 IU once a week by an MD. Apparently I had very low vitamin D, have darker skin, and live in a dark part of the world during winter. Which is more than was used in this experiment.

Usually, when I f-up my sleep schedule and start suffering from insomnia I go bush. Several days in wilderness in my swag and fully disconnected cures it like magic.

I have always wondered whether inducing exhaustion through exercise and strainous activities throughout the day could help overcome insomnia.

I thought so but it seems to be limited.

Swimming, jogging etc is amazing at relaxing, and making you want to sleep.

But there are times no amount will make you sleep.

For instance I bike commute daily, and yet work issues are making me sleepless.

> Don’t eat late at night.

Not eating late at night is the healthy choice but it definitely delays sleep onset for me. On the other hand I do seem to need less sleep if I don't eat late in the evening.

Eating later will cause reflux which will eventually cause you serious problems later in life. Initially reflux is noticeable, but after a while you will have it without any pain or spasms, and that is very bad.

The worst end result is cancer, and this is one of the worst forms of it to have.

How are you supposed to know when you fell asleep?

Interesting. It's a fixed point solution concept, where you do like a PD controller-style iterative re-adjustment based on the current sleep efficiency score. Just happens I was talking to my dad about fixed point approaches to sleep just a few hours ago.

When people change their wakeup time, one of two things happens: they either stick to their new habit successfully, or they break it, ignore it's broken, leave the alarm on but just ignore it. There's little opportunity to break free from the cycle because to re-enter the cycle is harder than exiting it. So you need some mechanism that provides sporadic reward for re-entering participation into the desired behavior.

Whatever your fixed point of behavior becomes, it becomes what you stick to. So you have to 'jam' the fixed point nature. Jostle it with some noise. Basically - change the meta.

My proposed solution models bedtime as a game, played by "night me", "morning me", and a randomized alarm clock.

  Alarm clock: either goes off at e.g. 5am (I want to wake up at 6am, so 5am is like a kind of punishment) with probability p, or it doesn't go off at all

  Night me: I won't know whether I will just 'sleep in' (relative to 5am), or whether I will be rudely awoken. The hope is it makes me hedge my bets - maybe I'd strategically adapt my nighttime activity, prudently leaning towards winding down earlier if I'm uncertain about my rest that night. Or I'll just be tired anyway because I will have repeatedly already demonstrated that I haven't yet learned the lesson to hedge one's time.. (ie got woken up because I got greedy about bedtime several times already) ..and go to sleep earlier anyway.. 

  Morning me: the subconscious me, really- I only have one move - to wake up - but in some sense it's out of my control until I've 'trained' myself via this routine. The score of this move is a function f(whether alarm actually went off, time I would have naturally woken up):

  - the alarm doesn't go off, I get to sleep in, yay, bc if I'm sleeping in I must have needed it

  - the alarm doesn't go off, I woke up early anyway, nice, that's the desired behavior

  - the alarm goes off, I don't get to sleep in, boo, I should have gone to sleep earlier, oh well, I was or will be tired tonight, anyway, probably - this is the only real 'negative' quadrant

  - the alarm goes off, but I was already awake around that time anyhow, meaning I've become 'over-trained' for my desired target - meaning, I can shut this whole game down for a few days!

By sort of anticipating a kind of homeostatic control that will make my wakeup time drift away from my overshoot target (5am) back to my actual target (6am), I will probably wake up later than this unusually punishing wakeup time. And, as a bonus, it will be perceived as sleeping in, simultaneously :)

It's gotta be that I could possibly get so wiped that I'd even be willing to go to bed not just like 30 minutes sooner but e.g. 3 hours sooner. That level of signal in behavior change is what I'm after, if possible

This is kind of like the "folk theorem" in game theory - you can kind of fudge the payoffs in infinite games Nash equilibria to basically be whatever you like. You only need to dish out as little punishment as is necessary to encourage positive reinforcement becoming the Nash equilibrium. So, in this context - pausing the 'game' a few days once I naturally overachieve my wakeup target = sleeping in more = higher ratio of the game payoffs being good ones = Nash equilibrium to pursue current mode of 'play'.

I hope by reaching a kind of a Bayesian Nash Equilibrium in this weird bedtime 'game', I jam my own meta, and avoid fixed point patterns where I just ignore my original goal. Though, to be honest, that's likely what will end up happening in the long term haha.

ps: you could also have a kind of stochastic bedtime approach to night time distractions too. Basically, somehow, set it up so that there's a p probability risk that you lose all your open tabs, unsaved files, internet connection, active phone calls, etc., at some uniformly unpredictable time e.g. up to 3 hours before bedtime. Hopefully you plan ahead, save all your work, and wind down far in advance, so you don't get burned by just surfing the net too long or something like that.

Reply on news.ycombinator.com