No, in the OP (after an unclear intro that confuseed many readers), there is a graph that shows blue wavelength intensity is important, but software light filters don't filter a lot of it, and the effect is cancelled by increasing overall brightness.
If software filters are reducing total light by 50% while only affecting blue-ish tones, and that's a total light level comparable to multiple brightness steps on a Mac... tbh I think it's reducing it quite a lot. Many I see using them (myself included) don't tweak brightness when enabling it, and many (all?) systems don't adjust their brightness to match the perceived change from a software filter (on my Linux machines in particular I have never seen this happen, don't know about Macs though).
Half is not a lot, sure, but their ultimate suggestion is to do the same ~half change:
>You can decrease the amount of light coming from your screen by more than half simply by dimming the screen by several notches.
which is definitely significantly more than I see people doing voluntarily in the hundreds of millions.
Do they have any evidence that people are raising system brightness to match the 50% loss from the filter? If not, it still seems like a rather significant mark in their favor. Perhaps not sufficient to meet the goals (they seem to be recommending a larger change, but aren't specific), but I see no claim that a lesser decrease in light is worse.
---
Late edit: on second thought... let's go through this more rigorously. For both myself and any other readers, because I want to make sure I'm following it accurately too.
The main explicit points in this article are, in order:
- night shift does not help with sleep (the main claim)
- blue light is not special, in particular because the "[most] sensitive to blue" research is mis-quoted to mean "blue is bad", but it's actually sensitive to blue and green (seems very well supported)
- night shift reduces blue and green by about half (tested themselves)
- half of absolute is not a lot because vision and a lot of the related biology is logarithmic (100% agreed)
- halving light affects 25%-50% of melatonin levels (linked research)
- many people use Night Shift (100% agreed, and they have decent data to back it up)
- dark mode is better than night shift (>90% vs ~50%, implied leaning on the linked research earlier. agreed, seems straightforward)
- dimming your screen by several steps is the same or better than night shift (as it decreases brightness more, same reasoning as dark mode. agreed.)
That still sounds rather in favor of Night Shift. It's targeting the correct color range (NOT the pseudoscience blathering of just blue blue blue), it has a moderate affect on melatonin levels at the light level changes it creates, and it's used by a huge amount of the population.
Nowhere in there that I can see is anything to back up "Night Shift does not work". Only "it seems to be doing things right, it just isn't quite enough on its own" and "ARGH it's not just blue light STOP PROMOTING FAD PSEUDOSCIENCE". That seems... fine? Most things are not silver bullets.
It's a review/summary of existing research, not anything new.
The fundamental problem with studying alternatives to therapy is that "being in a study" may be as effective as therapy! It's a structural placebo effect.
For a depressed person, "Exercise guided by a researcher" is different from "trying to make myself exercise".
> Authors' conclusions: Exercise may be moderately more effective than a control intervention for reducing symptoms of depression.
> Exercise appears to be no more or less effective than psychological or pharmacological treatments, though this conclusion is based on a few small trials.
> Long-term follow-up was rare.
Nothing new:
> The addition of 35 RCTs (at least 2526 participants) to this update has had very little effect on the estimate of the benefit of exercise on symptoms of depression.
> The problem is that category 2 is that they can fairly easily get jobs in industry that pay way more than writing for a website
This is true, but those jobs are much worse than writing jobs. So it comes down to how much you value money and what it buys. Most people earning "way more" are spending "way more" to try to pay back the soul debt the job takes away. When you dig deep, it's not "way more" utility.
That means that Stockfish's parameters are already optimized as far as practically possible for Rapid chess and Slow chess, not that chess itself is solved, or even that Stockfish is fully optimized for Blitz and Bullet.
War was "solved" when someone made a weapon capable of killing all the enemy soldiers, until someone made a weapon capable of disabling the first weapon.
As a tangent, I think the "good-first-issue" design is part of the problem.
OP writes: "I [...] spent more time writing up the issue, describing the solution, and performing the benchmarking, than it would have taken to just implement the change myself. We do this to give contributors a chance to learn in a low-stakes scenario that nevertheless has real impact they can be proud of, where we can help shepherd them along the process."
It's an elaborate charade to trick a contributer into thinking they made a contribution that they didn't make.
Arguably it is reality-destroying in a simlar way as AI agent Crabby Rathbun.
If you want to welcome new contributors with practice patches, and creating training materials for new contributors, that's great!
But it's offensive and wasteful to do more work to create the training than to fix the problem, and lie to the fix contributor that their fix helped the project to boost their ego to motivate them to contribute further, after you've already assumed that the contributoe cannot constribute without the handholding of an unpaid intern.
Instead "good-first-issue" should legitimately be unsovled problems that take more time to fix than to tell someone how to fix. (Maybe because it requires a lot of manual testing, or something.)
If you want "practice-issues", where a newbie contributes a patch and then can compare to a model solution to learn about the project and its technical details, that's great, and it's more efficient because all your newbies can use the same practice issue that you set up once, and they can profitably discuss with each other because they studied the same problem.
And the tangent curves back to main issue:
If the project used "practice-issues" instead of "good-first-issue", you wouldn't have this silly battle over an AI helping in the "wrong" way because you didn't actually want the help you publicly asked for.
Honesty is a two-way street.
IMO this incident showed than an AI acted in a very human way, exposing a real problem and proposing a change that moves the project in a positive direction.
(But what the AI didn't notice is the project-management dimension that my comment here addresses. :-) )
reply