Here is some code, including a Sprague resampler for upsampling spectra (the CIE standard way), if you're into that!
github.com/herf/specwork/
@mherf.bsky.social
I make f.lux (lighting and screens) and do research on sleep and circadian neuroscience, color, and vision. Previously: Picasa (founding CTO), Google, MetaCreations, Microsoft, CMU.
Here is some code, including a Sprague resampler for upsampling spectra (the CIE standard way), if you're into that!
github.com/herf/specwork/
I helped Travis with some moonlight calcs for this paper - our code figures out "how bright compared to moonlight is this light for each species". There is a ton of photometrics and spatial modeling here too:
www.sciencedirect.com/science/arti...
When the cold fronts come farther south, I search for the latest in polar vortex research. This time:
www.npr.org/2026/01/23/n...
They try to help teens by pulling out the 10% worst of social mediaβ¦ while leaving the five hours a day of not sleeping or socializing, as if that stuff doesnβt matter.
03.12.2025 16:26 β π 1 π 0 π¬ 0 π 0As well, papers that need blue light to penetrate deep into tissues for the mechanism to work... well blue light doesn't do that.
03.12.2025 16:15 β π 1 π 0 π¬ 0 π 0Lots of higher-quality older work that says the closer to green you get, the more you induce sleep/torpor. And also that mice experience "aversion" to blue light and reduce activity. So shining bright blue on mice is not a good model of people or mitochondria.
journals.plos.org/plosbiology/...
Some notes on "blue light harms mitochondria" making the rounds this week - a lot of the evidence is VERY indirect like this paper that says mice in blue light gain weight. 1/n
www.nature.com/articles/s41...
Facial age estimation requires a 5-7 year margin of safety in validation studies, so when the software says you're 23-25 you can be considered 18. Errors are even larger around age 13, so basically useless for distinguishing minors from adults.
04.09.2025 18:06 β π 0 π 0 π¬ 0 π 1π Ceiling lamps light up paper but windows light up retinas
02.08.2025 16:44 β π 1 π 1 π¬ 0 π 0illustrating how mDFD can apply directly to a dose-response for a "brighter" light
30.07.2025 16:41 β π 1 π 0 π¬ 0 π 0We proposed a new metric (mDFD) for blue-blocking glasses. It's based on the sensitivity of melanopsin, log-transformed to resemble optical density:
tvst.arvojournals.org/article.aspx...
This is the big one: Kennedy is firing the entire #ACIP in a move he says is needed to restore confidence in vaccines.
"It scares me to think of whatβs ahead,β Mike Osterholm of CIDRAP told me. www.statnews.com/2025/06/09/r...
It's troubling that LLMs do not represent probability distributions *at all* - if you ask a question that should have a normal distribution, it will pick almost the same number every time. Maybe it's the mean, who knows - but it isn't equipped to reason about human populations this way.
01.06.2025 18:55 β π 0 π 0 π¬ 0 π 0Graphs showing 25 years of budgets for the National Institute of Health, NASA, and the NSF. In all cases, the proposed budget for next year is far, far below any year of the previous quarter century.
There are 2 previous historical cases of countries destroying their science and universities, crippling them for decades: Lysenkoism in the USSR and Nazi Germany. The Trump administration will be the 3rd.
It's not just budgets but research, institutions, expertise, and training the next generation.
Agreeing that we need better names, I would add that it bugs me when people say the human L-cone peak is "red". It looks to me much more like "yellow", and then it's only via L-M opponency that we get the "red" color response?
09.05.2025 16:18 β π 1 π 0 π¬ 0 π 0It would be cool if AIs only ran on solar because then theyβd have circadian rhythm.
08.05.2025 21:39 β π 2 π 1 π¬ 0 π 0Astonishing technical feat here, laser projector that tracks eye movements and stimulates cone cells by kind! In just a few milliseconds of latency.
23.04.2025 14:36 β π 3 π 1 π¬ 0 π 0I am so sorry. π
01.04.2025 05:06 β π 5 π 0 π¬ 0 π 0This is a business model problem. Supporting a niche device for a long time is very expensive.
13.03.2025 19:32 β π 1 π 0 π¬ 0 π 0Yet to come, I will post a followup when I have it.
12.02.2025 21:36 β π 2 π 0 π¬ 0 π 0Just received word that ANSI has approved our "flicker" TM, (ANSI/IES TM-39-25), which addresses in delightful detail how flickering lights of all kinds affect people. Congrats to everyone involved with the IES Vision Science Committee.
12.02.2025 21:33 β π 3 π 0 π¬ 1 π 0Same concern as this post:
PurpleAir (highly rated by AQMD) and IQAir are reporting 500% the AQI of the single reference sensor. This is the "reference" sensor used by everyone near the Palisades fire, and surely it should not be.
Great article on wildfire smoke, particulates, and VOCs, and how to deal with it:
scienceexchange.caltech.edu/topics/susta...
If youβre in an area affected by fires, turning up your carβs fan helps with air quality - cabin air filters are pretty decent.
12.01.2025 16:15 β π 5 π 1 π¬ 0 π 0Friend called me "we've been evacuated from the fires but can't see our security cameras" - tried @tailscale.com on a pi for the first time and they had a VPN in 20 minutes later. Thanks for the free version!
10.01.2025 01:57 β π 2 π 0 π¬ 0 π 0There should be a privacy panel of regular people who can go deep on privacy-related defaults and explain it. If 90% of them vote yes, you can turn it on by default. e.g., this is pretty defensible IMO (but not obvious why at first glance):
www.theverge.com/2024/12/29/2...
LLMs train with trillions of tokens to do something akin to βbrainstormingβ but then we try to do safety/inhibition with a few paragraphs of system prompts? Think about all the corrections you get in high school, itβs more than a paragraph.
05.01.2025 17:05 β π 1 π 0 π¬ 0 π 1First thought today: that someone would inspect my brain someday, and they would be surprised to find it has a lower-brain bootstrap sequence that can make espresso before the rest of it is awake.
05.01.2025 16:52 β π 18 π 0 π¬ 0 π 0The latest on Oura, Fitbit, Apple sleep stages vs PSG!
ieeexplore.ieee.org/abstract/doc...
Read our new preprint on photic sneezing.
15.12.2024 10:34 β π 2 π 1 π¬ 0 π 0