Daniel's been thinking about the control stack behind the warm screen — not the "is blue light bad" question, but the actual levers under the hood. He's been a blue-light-filtering devotee for years, and two things have shifted his thinking. One: the emerging picture that for circadian effects, overall brightness is doing more work than the blue content of the spectrum. A bright warm screen can be worse than a dim cool one. Two: from lived experience, you can overshoot in the warm direction. Push the colour temperature down to deep amber and he finds it genuinely harder to concentrate — subtly sleep-inducing, which is fine at eleven p.m. and terrible at four p.m. There's a focus cost to over-warming that the "just install f.lux" discourse never mentions.
That focus cost is real and almost nobody talks about it.
So the real ask is the mechanism stack. What levers are we actually pulling under the hood, and which of them are real versus cosmetic. Gamma ramp manipulation — what f.lux, Redshift, GNOME Night Light, all of them actually do. True backlight control and why minimum brightness is still too bright. DDC/CI for multi-monitor desktop setups — Daniel's situation. Hardware low-blue-light modes in monitor firmware, and whether those are spectrally different from a software ramp. Then the payoff: sane parameters for a defensible circadian desktop setup, not a superstitious one.
And he wants us to be honest that this is contested rather than settled. The brightness-over-wavelength reading has experimental support, but expert consensus still holds that melanopic content has independent effects at matched intensity. Both readings point the same practical direction: dim first, warm second.
Let's start with the gamma ramp, because that's what most people are actually using and it's the layer where the most confusion lives.
What f.lux and Redshift and Night Light actually do is rewrite the GPU's colour lookup table. Every pixel your graphics card wants to display has a red, green, and blue value — each an eight-bit number, zero to two hundred fifty-five. Before those values hit the display, they pass through three lookup tables, one per channel. The LUT maps each input value to an output value. Normally it's a straight line — input fifty maps to output fifty. What the night-mode tools do is bend those curves down for the blue and green channels. So when the framebuffer says "display blue at intensity one hundred," the LUT says "actually, make that sixty."
And the backlight doesn't change.
The backlight doesn't change at all. This is the crucial thing. A white-LED backlight is a blue LED — typically with a peak around four hundred fifty nanometres — coated with a yellow phosphor. That blue pump spike is always there, at full power, regardless of what the LUT is doing. What the gamma ramp changes is how many subpixels are being asked to pass that blue light through. You're attenuating the signal, not the source.
So the blue photons are still being produced. You're just blocking more of them at the LCD layer.
Right. The question is how much that actually reduces melanopic stimulus at the eye. And the answer is... it depends on the display and how aggressive the ramp is, but you can get meaningful reduction. The ipRGCs — the intrinsically photosensitive retinal ganglion cells that drive melatonin suppression — they're most sensitive around four hundred eighty nanometres. That's not exactly the LED blue pump at four fifty, but there's overlap. Attenuating the blue channel in the LUT does reduce the total radiant power in that region. The problem is you're doing it by throwing away bit depth.
Explain the bit-depth cost.
If your blue channel originally had two hundred fifty-six possible values and you're now mapping the top half of the range into a compressed space, you've lost effective bit depth. You get banding in smooth gradients. Colour accuracy is wrecked — you literally cannot display certain colours anymore because the blue subpixels can't reach their full intensity. For reading text or coding, fine. For anything where colour matters, you're working with a crippled display.
And Daniel's point about the backlight spike at four fifty nanometres still being there — does that mean some blue light is leaking through regardless?
Some, yes. The LCD layer isn't a perfect shutter. Even a "black" pixel passes some light. So you've got this baseline leakage of the blue pump through every subpixel regardless of what the LUT says. The gamma ramp reduces the modulated portion of blue light — the part that varies with image content — but the backlight's spectral peak is still present in the leakage. How much that matters physiologically... honestly, I don't think anyone has measured it directly for LCD panels in a circadian context.
So the gamma ramp is a real lever but an imprecise one. What about the second layer — actual backlight control?
This is the big one. The luminance lever. And it's where the "dim first" part of the heuristic lives. The Scientific Reports paper from January — Terán and colleagues — they measured melatonin suppression values across fifty-two lamps and normalized everything to the same photopic illuminance, about forty lux, equivalent to a seventy-five-watt incandescent at two metres. At matched brightness, cool LEDs suppressed melatonin at about twelve point three percent, warm LEDs at three point six percent, incandescent at one point five percent. So spectrum matters. But brightness is the multiplier on all of it. A cool LED dimmed to half the illuminance is going to deliver less melanopic dose than a warm LED at full brightness.
And the consensus survey from twenty twenty-three — the Frontiers in Photonics paper — they found over ninety percent of experts agreed that increasing indoor light intensity at night increases circadian disruption and suppresses melatonin. Both levers are real.
Both real, and brightness is the one with the bigger dynamic range. The problem Daniel's getting at is that most panels won't go low enough. The minimum backlight setting on a typical desktop monitor is still pumping out maybe fifty to eighty nits. In a dark room, that's a lot. Your phone can go down to one or two nits because it's designed for dark bedrooms. Desktop monitors aren't.
And then there's the PWM problem.
Pulse-width modulation. To get brightness below the maximum, most LED backlights don't actually reduce the current — they just turn the LEDs on and off really fast. At full brightness, they're on continuously. At fifty percent, they're on half the time. The frequency varies wildly — some panels do two hundred hertz, some do several thousand. Below a few hundred hertz, a subset of people perceive flicker. It causes eye strain, headaches. Even people who don't consciously see it can have their saccadic eye movements disrupted.
So you dim the backlight to protect your circadian rhythm and trade it for a headache.
That's the trade. DC dimming — actually reducing the current rather than pulsing — exists but it's less common because it can cause colour shift at low currents. The LEDs don't emit exactly the same spectrum when under-driven. So most manufacturers went with PWM.
What about the software dimming overlay trick? The thing where you put a translucent black overlay on top of everything?
That's multiplying the framebuffer values down further — it's just another layer of the same gamma ramp approach. You're not reducing the backlight at all. The backlight is still at whatever brightness it's set to, and now you're asking the LCD to block even more of it. You get even more bit-depth crush, even worse contrast. It works in the sense that less light reaches your eye, but it's the worst way to achieve that.
Which brings us to layer three — DDC/CI. This is Daniel's actual situation. Multi-monitor desktop, no laptop brightness keys. He's using ddcutil to talk to the monitor firmware over the I2C bus.
DDC/CI — Display Data Channel Command Interface. It's a protocol that lets your computer send commands to the monitor over the same I2C lines that carry the EDID data. The monitor exposes a set of virtual control codes — brightness, contrast, colour presets, input source, sometimes red-green-blue gain. The standard is VESA MCCS, Monitor Control Command Set.
And it's flaky.
It's incredibly flaky. The standard defines the commands, but implementation is up to the manufacturer and they are not diligent about it. Some monitors only respond to brightness. Some expose contrast but changing it does nothing. Some accept the command and then ignore it until you power-cycle. Some work perfectly for weeks and then stop responding until you unplug and replug the display cable. The I2C bus can get wedged. Multiple monitors on the same bus can conflict — ddcutil has to handle bus enumeration and sometimes the mapping between I2C bus numbers and physical displays shifts between reboots.
So you can script it but you need error handling.
You need error handling and you need to test it on your actual hardware. The basic loop is: enumerate the displays, for each one set brightness to a target value, verify it took, log failures. You wire that to a cron job or a systemd timer that fires at sunset and again at a scheduled bedtime. Daniel's on Linux, so ddcutil is the right tool — it's actively maintained, it handles the I2C bus probing, it can dump all the capabilities a monitor advertises. But it's not plug-and-play. You will spend an afternoon getting it working.
And once it works, you have real backlight control across multiple displays from a script. That's the "dim first" lever, automated.
And it's worth doing because it's the only software-accessible path to actually reducing the backlight output on a desktop monitor. Everything else is just asking the LCD to block more light from a backlight that's still running at full power.
Layer four — the hardware low-blue-light modes. The "Reader" or "Low Blue Light" preset in the monitor's on-screen menu. Is that doing something spectrally different from the gamma ramp?
It depends on the monitor, and this is where panel technology really matters. On a standard white-LED backlit LCD, the Low Blue Light mode is typically doing the same thing as the gamma ramp — attenuating the blue channel — but it's doing it in the monitor's internal image processing pipeline rather than in the GPU. The backlight spectrum is unchanged. You're still dealing with that blue pump at four fifty nanometres.
So it's the same thing one layer down.
Same thing, but with one potential advantage: the monitor's internal processing might use a higher bit-depth internally before dithering down to the panel's native bit depth, so you might get less banding than a GPU-side gamma ramp. Might. It depends entirely on the scaler hardware.
What about OLED and QD-OLED?
Different story. OLED panels don't have a backlight — each pixel is its own light source. The emission spectrum of an OLED is fundamentally different from a white-LED backlight. There's no blue pump spike at four fifty nanometres — the blue subpixels emit in a broader peak. QD-OLED uses blue OLED material as the base emitter with quantum dots converting some of that to red and green. The blue peak is still there but it's different in shape and intensity. The Terán paper's spectral analysis is all on LED, CFL, and incandescent lamps — they didn't measure OLED panels. But the principle holds: if you can shift the actual emission spectrum, you're doing something the gamma ramp can't.
The hardware modes on some high-end monitors do exactly that — they have multiple backlight LED types or they adjust the white point at the firmware level by changing the relative drive currents to different LED strings. That's a real spectral change. But on a consumer desktop monitor with a standard white-LED backlight, the Reader mode is almost certainly just blue channel attenuation in the scaler.
So the layer cake, from most cosmetic to most real: gamma ramp is attenuating the signal, hardware preset is the same thing in a different place, DDC/CI lets you control the actual backlight brightness, and a true spectral shift requires different hardware entirely.
That's the stack. And Daniel's right that most people are only pulling the first lever and thinking they've solved the problem.
Let's talk about the over-warming problem. This is the part nobody discusses.
The cognitive cost of pushing colour temperature too far toward amber during working hours. Daniel's describing something real. The circadian system interprets warm, dim light as an evening signal. The ipRGCs project to the suprachiasmatic nucleus, which drives melatonin release from the pineal gland. But they also project to a bunch of other brain regions — the olivary pretectal nucleus for pupil control, the lateral geniculate for visual processing, and areas involved in alertness and attention. When you bathe your visual field in deep amber light, you're sending a strong "nighttime" signal to a system that's been tuned by millions of years of evolution to respond to exactly that cue.
The response is "wind down."
The response is wind down. Reduced alertness, reduced cognitive throughput. There isn't a clean dose-response curve in the literature for colour temperature versus cognitive performance during daytime, because most of the research is focused on evening exposure and sleep outcomes. But the consensus survey found that increasing indoor light intensity during the day — within the fifty to five hundred lux range — enhances circadian entrainment and improves daytime alertness, with about seventy-five percent expert agreement. By extension, if you're simulating evening light during the day, you're working against that.
What's the floor? What colour temperature should you not go below during working hours?
There's no hard number from the literature. The consensus survey notes that blue-enriched light during the day enhances alertness and mood — that's cool white, five thousand to sixty-five hundred Kelvin. For evening, the Terán data shows a fixed warm LED at about three thousand Kelvin reduces melatonin suppression by over threefold compared to cool LED. A tunable lamp at its warmest setting — twenty-one hundred Kelvin — can reduce suppression to near zero, point one percent. That's the sleep-promoting range.
Twenty-one hundred Kelvin at four p.m. is a bad idea.
It's a terrible idea if you need to focus. The focus cost Daniel's describing probably kicks in somewhere below about thirty-five hundred Kelvin for most people during daytime. That's a rough estimate — individual variation is huge, and the consensus survey explicitly reached consensus on that point: over ninety percent of experts agreed there's significant variation in individual sensitivity and circadian lighting should be personalized.
Which means you have to experiment on yourself.
You have to experiment. The sane approach is to keep daytime colour temperature at or above four thousand Kelvin — basically, don't warm your screen during working hours at all, or warm it only slightly. Start the shift maybe two hours before your target bedtime. The thirty-to-ninety-minute pre-sleep window is where the melatonin suppression model in the Terán paper is calibrated — that's the critical intervention window.
And brightness?
During the day, bright is good — it reinforces the circadian signal. In the evening, dim first. The Terán paper normalized to about forty lux at the eye, which is roughly what you'd get from a seventy-five-watt incandescent at two metres. A desktop monitor at typical viewing distance with the backlight at minimum might still be delivering more than that to your eyes. So dim it as low as you can while still being able to work. If you can't get it low enough with the backlight alone, then you add a mild warm shift on top.
The "dim first, warm second" heuristic in practice.
Dim first because brightness is the multiplier on everything. Warm second because at matched brightness, spectrum still matters — the consensus survey found over seventy percent of experts agreed blue-enriched light in the three hours before bedtime disrupts sleep more than blue-depleted light at the same intensity. The effect is real, it's just smaller than the brightness effect.
The blue-light-blocking glasses meta-analysis Daniel wanted us to look at — the Luna-Rangel paper from Frontiers in Neurology, November twenty twenty-five — that's a useful reality check.
Three double-blind crossover randomized controlled trials. Total sample size: forty-nine people. That's it. That's the entire high-quality evidence base for blue-light-blocking glasses measured with actigraphy — objective wrist-worn sleep tracking, not self-report. The pooled results: sleep onset latency improved by about five minutes, not statistically significant. Total sleep time increased by about nine minutes, not significant. Sleep efficiency and wake after sleep onset — no meaningful difference. The effect directions favoured the glasses, but none of it crossed statistical significance.
Forty-nine people across three studies.
The heterogeneity was zero — which means the studies all found similarly small effects, it's not that they disagreed with each other. The effect is probably real but small. The authors conclude these might help in clinical subgroups with circadian misalignment, not in the general population. Which lines up with the broader picture: if you're already sleep-deprived or have insomnia, maybe you're more sensitive to evening light. If you're a healthy adult with decent sleep, the glasses probably don't do much that dimming your lights wouldn't also do.
The Campo-Beamud review from June — the one in Archivos de la Sociedad Española de Oftalmología — that's even more blunt.
They reviewed the literature from two thousand nine to twenty twenty-five and concluded blue-filtering lenses give no clinically meaningful improvement in visual acuity or digital eye strain, and only small, inconsistent sleep effects. Their recommendation: light hygiene and reduced nocturnal exposure over filters. Not "buy these glasses." Just... turn down the lights.
Which brings us back to DDC/CI and actually turning down the lights.
The ambient room light question is the one that makes all of this either matter or not matter. The Terán paper's entire premise is that ambient room lighting is a significant source of circadian disruption. If your room is lit by a cool white LED ceiling fixture, that's delivering a melatonin suppression value of around twelve percent — and it's bathing your entire visual field, not just the screen-sized solid angle. Your phone's night mode is irrelevant in that context. The screen could be perfectly amber and you're still getting a full circadian dose from the room.
The exercise of filtering the screen is beside the point if the room is a cool white LED.
Completely beside the point. The sane setup is: address room lighting first. Warm bulbs in the evening — ideally tunable ones that can shift from cool during the day to twenty-one hundred Kelvin at night. The Terán paper showed tunable LED lamps can drop melatonin suppression from about ten percent at fifty-seven hundred Kelvin to point one percent at twenty-one hundred Kelvin. That's a hundredfold reduction. Then address the screen with the dim-first-warm-second approach. Then, if you still want to push further, the screen's gamma ramp or hardware preset is the fine-tuning knob.
The timing? When do you start the shift?
The melatonin suppression model is calibrated on a thirty-to-ninety-minute exposure before sleep. That's the window where light has its maximum effect on that night's melatonin onset. So the minimum defensible approach is: start dimming and warming ninety minutes before bed. If you want to be more conservative, two to three hours. The consensus survey framed the three-hour pre-bedtime window as the relevant period for blue-enriched light disruption.
What about the case for not warming at all during the day? Daniel mentioned he's found the focus cost real.
The case is strong. Daylight is fifty-six hundred to sixty-five hundred Kelvin. Your brain interprets that as "active phase." The consensus survey found that blue-enriched light during the day enhances alertness and mood — that's not controversial. So during working hours, keep your displays at native colour temperature. Maybe sixty-five hundred Kelvin if that's what they're calibrated to. Don't touch the gamma ramp. Use DDC/CI to set brightness high enough to be comfortable — probably somewhere between one hundred twenty and two hundred nits depending on your room lighting — and leave it there. When the sun sets, or two to three hours before bed, trigger the shift: dim the backlights first, then warm the colour temperature.
The floor for the warm shift — where do you stop?
For evening work where you still need to focus, I'd say don't go below about thirty-four hundred Kelvin. That's roughly the colour of a halogen lamp — warm but not amber. For the last thirty to sixty minutes before bed, if you're just reading or winding down, you can push lower — twenty-five hundred, even twenty-one hundred. But Daniel's experience is the guide here: if you feel your focus slipping, you've gone too far. Back it off.
The defensible setup for a desktop workstation: DDC/CI scripts controlling backlight brightness on a schedule. Room lighting dealt with first — tunable warm bulbs for evening. Daytime: native colour temperature, bright enough to work comfortably. Two to three hours before bed: dim the backlights. Ninety minutes before bed: add a mild warm shift, maybe thirty-four hundred Kelvin. Last thirty minutes: push lower if you want, but only if you're done focusing.
Accept that this is all probabilistic. The effect sizes are small at the individual level. The blue-light-blocking glasses -analysis found a five-minute improvement in sleep onset that wasn't statistically significant. You're not going to transform your sleep by tweaking your monitor settings. But the cumulative effect across room lighting, screen brightness, screen spectrum, and timing — that probably matters. And the cost of getting it wrong is basically zero, as long as you're not over-warming during the day and sabotaging your focus.
On the record. Predictions.
Within two years, a major monitor manufacturer will ship a desktop display with a genuine circadian mode — not a blue-light preset that's just a gamma tweak, but a backlight with independently addressable cool and warm LED strings, controllable over DDC/CI, with published spectral data. The demand is there and the LED cost is negligible.
I'll go narrower. Within eighteen months, someone will publish a study measuring actual melanopic dose at the eye from a desktop monitor with gamma-only warming versus true backlight spectral shift, and the difference will be large enough to embarrass the "just use Night Shift" advice.
That would be useful.
Thanks to Hilbert Flumingtop for producing. This has been My Weird Prompts. Find us at my weird prompts dot com or email the show at show at my weird prompts dot com. We'll be back soon.
