Daniel's been building a circadian lighting plugin for Home Assistant, three states: daytime, warm low-Kelvin, red before bed. And he's stuck on the scheduling question. Absolute offset — red light three hours after sunset — versus proportional, red light at thirty percent of the way through the night. He's come round to thinking proportional is the more elegant of the two, and he wants us to argue him out of it if he's wrong. But he also wants the bigger picture. Has anyone defined golden hour as a fraction of daylight? What are the natural percentage phases? What's actually been shipped in circadian lighting, in Adaptive Lighting style integrations, in commercial and architectural work, in the research literature? And where does chronobiology actually put the inflection points, as opposed to where designers assume they are?
And the latitude question. He named it himself. Fifty-three degrees north, a proportional night is a wildly different length in June and December. Thirty-two degrees, it barely moves. So the parameterization question changes depending on where you stand on the planet.
Right. And I think that's the thing that makes this worth an episode rather than a forum post. It's not just a scheduling API question. It's about whether human biology runs on clock time or on proportional time, and whether nature gives us a ruler or we're just choosing which ruler to use.
So before we pick a ruler, we should ask what we're actually trying to measure. Because the thing that jumped out at me reading Daniel's prompt is that both of his candidates are anchored to sunset. And sunset is a free anchor, trivially derivable from geolocation, adapts to season and latitude for free, as he said. But the question is whether sunset is the thing that matters. And chronobiology says... no. Not really.
Say more.
The two-process model of sleep regulation. Process S is sleep pressure, it builds with wakefulness, it's homeostatic. Process C is the circadian drive, the internal clock. The evening transition that lighting designers are trying to support is the melatonin rise. Dim light melatonin onset, DLMO. And DLMO typically occurs two to three hours before habitual sleep time. Not two to three hours after sunset. Two to three hours before you habitually go to bed.
So the biological inflection point is sleep-anchored, not sun-anchored.
And that's the first crack in the proportional scheme. Daniel's instinct is that proportional is more elegant because it keeps the phase in a consistent relative position over the calendar year. But consistent relative to what? If the thing you're actually trying to support is the melatonin rise, then the anchor should be habitual sleep time, and the relationship between sleep time and sunset is loose. People go to bed at roughly the same clock time all year, regardless of when the sun sets. Dinner at seven is dinner at seven whether the sun set at four or eight.
So neither parameterization is biologically grounded. Absolute offset from sunset is wrong because sunset isn't the anchor. Proportional through the night is wrong for the same reason. They're both approximations of an anchor we can't directly observe.
And the phase response curve makes this even sharper. The PRC to light is the key mechanism here. Light exposure in the evening delays the circadian clock. Morning light advances it. So the timing of when you switch to warm or red light isn't cosmetic. It's a phase-shifting intervention. You're not just making the room feel cozy, you're nudging the body's internal clock. And the PRC is anchored to internal circadian phase, not to sunset. Internal phase drifts relative to external time. It varies by chronotype, by age, by individual.
So the elegant proportional scheme is elegant in the way a perfectly drawn map of the wrong territory is elegant.
That's the thing. The designer assumption problem. Most circadian lighting schemes assume the evening transition should track sunset. But the relevant anchor is the individual's sleep-wake schedule. And here's the uncomfortable part: the circadian nadir, the lowest core body temperature, sits roughly two hours before habitual wake time. So the morning inflection point is also sleep-anchored. Not sunrise-anchored. The body's two big daily breakpoints, melatonin onset in the evening and temperature nadir in the morning, they're both tied to when you sleep and when you wake, not to where the sun is.
Which means a fixed proportional schedule is a one-size-fits-all solution to a problem that is individual. Phase angle of entrainment — the relationship between internal circadian phase and the external light-dark cycle — varies by chronotype. A night owl's DLMO is later relative to clock time than a morning lark's. So even if you got the proportional math perfect for one person, you'd be off for the next.
And this is where I want to be careful, because I think there's a version of Daniel's instinct that survives this. The body does entrain to light. The light-dark cycle is the primary zeitgeber, the time-giver. So the sun isn't irrelevant. But the entrainment is mediated through the phase response curve, which means the body responds to light at specific phases of its own internal cycle. It's not a simple photocell. It's not "sun is down, release melatonin." It's a dynamical system with its own inertia.
So if we're honest, the ground truth is that neither absolute nor proportional-to-sunset is biologically correct. Both are proxies. And the question becomes which proxy is less wrong for the actual use case.
Right. And I think that's the frame we should carry through the rest of this, because Daniel asked a bunch of specific questions and we should actually answer them. The golden hour question. Has anyone defined golden hour as a fraction of daylight?
And the answer there is pretty clear. No. Golden hour in photography and cinematography is defined by solar elevation angle, not by fraction of day. It's the period shortly after sunrise and before sunset when the sun is low, between roughly negative four and positive six degrees of elevation. And it's typically described in absolute minutes. Roughly thirty to sixty minutes, depending on latitude and time of year. The visual quality depends on the angle of the sun through the atmosphere, not on what fraction of the day has elapsed.
Which is a nice example of the same tension. Photographers didn't choose absolute minutes because they were being unsophisticated. They chose it because the thing that matters, the quality of the light, is a function of solar geometry, not of the interval between sunrise and sunset. A proportional golden hour would be wrong at the equator and wrong at fifty-three degrees north, because the sun drops through those low elevation angles at very different speeds.
And the ancient timekeeping thing. Daniel realized afterwards that proportional phases are more or less what temporal hours already did. And he's right. The classical and medieval world divided daylight and night into twelve equal parts each. Seasonal hours, unequal hours. An hour of daylight in summer was longer than an hour of daylight in winter. The sha'ah zmanit in Jewish law, the Roman and Babylonian systems, canonical hours for prayer, all of it. It persisted for millennia because it tracked lived experience of daylight rather than abstract clock time.
And the medieval monks would have recognized Daniel's proportional scheme instantly. But they weren't trying to do it at fifty-three degrees north in December with electric lights. The temporal hour system was designed for a world where daylight was the thing that structured activity. When night is sixteen hours long, a twelfth of the night is a very long hour. The system starts to feel wrong precisely when the thing it's proportional to stops matching human behavior.
What have people actually shipped? Daniel asked about prior art. Adaptive Lighting, the popular Home Assistant integration, uses absolute offsets from sunset and sunrise. Typically a fixed duration for the transition window. Sunset offset, sunrise offset, transition length.
The circadian lighting community has largely converged on that. Absolute offsets with configurable transition lengths. It's not that nobody thought of proportional. It's that absolute offsets won the argument in practice. And I think the reason is exactly what we've been circling. The evening transition is socially anchored. Dinner, wind-down, bedtime. Those happen on clock time. The dawn transition is different.
That's the hybrid synthesis, isn't it? Proportional for the dawn transition, because dawn tracks the light people wake to, and absolute or socially anchored for the evening, because dinner at seven is dinner at seven.
The research on social jetlag and meal timing would predict exactly that. Social jetlag is the mismatch between your biological clock and your social clock. On weekends people sleep later, which shifts their internal phase, and Monday feels like jetlag. The evening is where social and biological clocks collide most. So if you're going to use a sun-anchored proportional scheme anywhere, morning is the place.
Let's make the latitude problem concrete, because Daniel asked about fifty-three versus thirty-two. At fifty-three degrees north, roughly the UK or Alberta, night length at the winter solstice is about sixteen hours. At the summer solstice, about seven and a half. So thirty percent through the night means red light at roughly two in the morning in December, but around ten at night in June. That's a six-hour swing in when the red light kicks in.
At thirty-two degrees north, roughly San Diego or Jerusalem, night only swings between about ten and fourteen hours. So the proportional and absolute approaches barely diverge. Thirty percent of a ten-hour night is three hours after sunset. Thirty percent of a fourteen-hour night is a bit over four hours. The difference is an hour and change. At fifty-three degrees, the difference is enormous.
That's the strongest case against proportional. At extreme latitudes, proportional scheduling produces behavior that feels wrong. In midwinter at fifty-three degrees north, a proportional evening schedule would push warm light absurdly early relative to when people actually eat dinner or wind down. Because the night is so long, a fraction of it is a huge chunk of clock time. Absolute offsets keep the transition anchored to social and behavioral rhythms that don't scale with the sun.
Then there's the commercial and architectural world. What do they actually ship? The big lighting control systems, the Lutrons and the rest, they've largely settled on schedules. Clock-based schedules, sometimes with an astronomical offset. But the trend in the last few years has been toward what's called human-centric lighting, and the more sophisticated implementations are starting to use personalization. Wearables, sleep trackers, that kind of thing. The frontier isn't sun-anchored at all. It's body-anchored.
Which brings us back to the phase response curve. If you had a wearable that could estimate your DLMO, your internal phase, you wouldn't need to choose between absolute and proportional. You'd just schedule the transition relative to your own body. But that's not what Daniel's building. He's building a Home Assistant plugin that has to work for other people, with just geolocation and a clock.
The practical question. Which parameterization would you ship if you were writing this for other people to use? And I think the honest answer is: both, but for different transitions. Absolute offset for the evening, because evening is socially anchored. Proportional for the dawn, because dawn is sun-anchored. And maybe expose the choice, because at thirty-two degrees north it barely matters and at fifty-three degrees north the user is going to have strong feelings.
The user at fifty-three degrees north in December is going to look at a proportional evening schedule and think the plugin is broken. Red light at two in the morning? That's not what I want. I want red light when I'm actually going to bed. And that's the failure mode. Elegance is a property of the parameterization. Correctness is a property of the relationship between the parameterization and the human.
I think that's the deepest point here. The parameterization question opens onto a deeper question about what we think lighting is for. Is it for tracking the sun, or is it for supporting the body, or is it for matching social rhythms? Those are three different targets. And they pull in different directions.
Where does that leave Daniel's instinct? I think the case for proportional is real. It's more elegant as a mathematical object. It adapts to season and latitude for free, exactly as he said. And the temporal hours precedent shows it's not a new idea. But the case against is that the thing it's proportional to, the night interval, is not the thing the body actually cares about. The body cares about its own internal phase, which is anchored to sleep and wake, and about social time, which is anchored to the clock. The night interval is a proxy for both and a poor one.
If you're going to ship a proxy, you should ship the proxy that least surprises the user. Absolute offsets from sunset are less surprising because they track clock time more closely. Proportional through the night is more surprising because it amplifies latitude and season in ways that don't match lived experience.
Let me put it in terms of the failure pattern. Absolute offset fails when the user's sleep schedule drifts relative to sunset. But that drift is usually slow, and the user can adjust the offset. Proportional fails when the night interval changes dramatically, which happens fast at high latitude, and the user has to keep adjusting the fraction, which is unintuitive. The absolute offset fails gracefully. The proportional scheme fails spectacularly.
There's a second failure pattern for proportional that doesn't get talked about. The fraction itself is hard to reason about. If I tell you "red light at thirty percent through the night," can you tell me what time that is? If I tell you "red light three hours after sunset," you know exactly what time that is. The absolute offset is legible. The proportional fraction is elegant but opaque.
That's the thing about elegance. It's a property of the system as seen from above. Legibility is a property of the system as seen from inside. And users live inside.
If I were shipping this, and Daniel asked, I'd ship absolute offsets for the evening transition, proportional for the dawn transition, and I'd document why. The evening is social. The dawn is solar. And I'd expose both parameterizations as options, because the latitude problem is real and the user at fifty-three degrees north is going to want to override whatever default I pick.
I'd add one more thing. A fixed clock option. Because Hilbert's about to tell us about a place where both parameterizations collapse entirely.
Hilbert: They do.
Hilbert: I spent a winter in Tromsø.
Hilbert: Night auditor at a hotel. The sun doesn't rise for six weeks. There's no sunset to anchor to. And the night interval is the whole day. So both of Daniel's schemes just... stop. Absolute offset from sunset, there's no sunset. Proportional through the night, the night is twenty-four hours. Thirty percent of twenty-four hours is seven in the morning. That's not a lighting schedule, that's just a clock.
Hilbert: The hotel ran on a fixed schedule. Full-spectrum bright light at eight in the morning, every morning, whether the sun was up or not. We called it aggressive breakfast lighting. It was printed on a card in every room. The card didn't change all winter. Nobody looked at the sun. The sun wasn't there.
Hilbert: Everyone's circadian rhythms went haywire anyway. The lighting helped, but it wasn't sun-tracking. It was social. Breakfast at eight, bright light at eight, because that's when people needed to be awake. The sun had nothing to do with it.
The third option. Neither absolute nor proportional. Just a fixed clock.
That's the thing the whole discussion had been missing. We assumed there's a meaningful day-night cycle to parameterize against. Tromsø in December says no. And the hotel's solution, socially anchored fixed schedule, worked better than any sun-tracking scheme. Because the sun wasn't the thing people needed to track. They needed to track breakfast.
That reframes the whole question. The parameterization choice assumes the sun is the right anchor. But maybe the right anchor is the social schedule, and the sun is just a convenient proxy for it in places where the sun actually shows up.
Hilbert: The card didn't change. That's the part I keep thinking about. All winter, same card. Nobody complained about the lighting. They complained about the dark, but not the lighting. The lighting was the one thing that worked.
Because it was honest. It didn't pretend to track the sun. It just said: this is when we wake you up, this is when we dim the lights, deal with it. And the body responded to the consistency, not to the solar accuracy.
That's the design provocation, isn't it? Maybe the most elegant parameterization is the one that admits it's a proxy and lets users pick which proxy they trust. Sunset, clock, or their own habits.
Hilbert: Or their own breakfast.
Hilbert: I'd have shipped the card.
The card. Fixed schedule, printed, no parameters at all. There's something to that.
Where does that leave Daniel's question? I think the answer is: ship absolute for the evening, proportional for the dawn, expose both, and document that at extreme latitudes the whole framing breaks down and the user should just pick a clock time. But the deeper answer is that the parameterization question is really a question about what you think the light is for. And the biology says the body wants light based on its own internal phase, which is neither clock time nor sun time.
That's the uncomfortable implication. The phase response curve says circadian lighting is actually about manipulating internal phase, not matching the sun. It's a phase-shifting intervention. The sun is a proxy. The clock is a proxy. The body is the thing you're actually trying to move. And the body doesn't care about your elegant fraction.
Which is why I keep coming back to the phase angle of entrainment. It varies by chronotype and age. A teenager's DLMO is later than a fifty-year-old's. So even the hybrid scheme, absolute evening, proportional dawn, is a population-level compromise. It's the best you can do with just geolocation and a clock. But it's not the truth.
Maybe that's the honest thing to tell users. Here's a reasonable default. Here's the knob. If it feels wrong, turn the knob. Because the thing the knob is a proxy for, your internal phase, is something we can't measure from a light bulb.
Daniel's proportional instinct is beautiful. It's the temporal hours insight, rediscovered independently. And it's wrong for the evening at high latitude, because the evening is social, not solar. But it's right for the dawn, because the dawn is solar, not social. And the whole thing collapses in Tromsø, where neither the sun nor the social schedule agrees to be the anchor.
The cutting-room floor detail I wanted to mention: the dim light melatonin onset literature actually shows that DLMO is more stable when measured relative to habitual sleep time than when measured relative to sunset. Across studies, the standard deviation of DLMO relative to sleep time is on the order of an hour. Relative to sunset, it's much wider. So the biology really is sleep-anchored, not sun-anchored. The data's been there for decades.
The open question I'm left with: is the natural parameterization actually a proxy for something else? Namely, that we want lighting to track either the sun or our social schedules, and the choice of ruler depends on which one we think matters more for a given transition. And chronobiology suggests the honest answer is that neither sun-anchored scheme is biologically optimal. The body wants light based on its own internal phase. So maybe the most elegant parameterization is the one that admits it's a proxy and lets users pick which proxy they trust.
Sunset, clock, or their own habits. Or their own breakfast, if they're in Tromsø.
Thanks to our producer, Hilbert Flumingtop.
This has been My Weird Prompts. Email us at show at my weird prompts dot com.
We'll be back soon.
