Eulogy for Dark Sky, a Data Visualization Masterpiece (2023)
Key topics
As the weather app world mourns the loss of Dark Sky, a masterpiece of data visualization, commenters are sharing their favorite alternatives and reminiscing about Dark Sky's innovative features. Some are raving about Meteoswiss, a Swiss weather app praised for its accuracy and design, while others are nostalgic for Dark Sky's intuitive UI and hyperlocal forecasts. The discussion reveals a consensus on Dark Sky's excellence, with many users fondly recalling its ability to predict rain with minute-by-minute precision. With weather apps remaining an essential part of daily life, this thread is a timely tribute to a pioneering app that raised the bar for weather forecasting.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
36m
Peak period
62
3-6h
Avg / period
14.5
Based on 160 loaded comments
Key moments
- 01Story posted
Jan 10, 2026 at 7:23 AM EST
1d ago
Step 01 - 02First comment
Jan 10, 2026 at 7:59 AM EST
36m after posting
Step 02 - 03Peak activity
62 comments in 3-6h
Hottest window of the conversation
Step 03 - 04Latest activity
Jan 11, 2026 at 6:49 PM EST
5h ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
The information design argument is 100% valid, but I also marvel that, having bought the company, Apple's weather app still isn't as precise or accurate. I don't know whether Apple's privacy focus prevents them making the same precise predictions, or if there is some other reason they don't, but it's sad that in 2025 we don't have the same level of performance as we did twelve years ago.
So much of weather forecasting, at that time, was about trends and probabilities. DarkSky was about events, certainty, and action.
It was truly ahead of anything else and forced a new standard.
But iOS has this now. It's the same thing. They integrated it from Dark Sky.
Is it not? The rainfall-per-minute over the next hour on iOS seems about the same accuracy as Dark Sky had -- I used Dark Sky for years. It wasn't perfect but it worked well enough, same as iOS did after. You can even scrub the precipitation map predictions and they look the same to me.
I know the Dark Sky prediction accuracy was greatly dependent on where you lived -- this is something that was widely discussed back in the day. If you've seen a drop in accuracy, did you simply move?
And just anecdotally, Dark Sky was a delight to use. Apple Maps makes it a chore to extract the same utility from their app.
(And Apple notifications are a mess generally. I constantly have notifications for something yesterday only show up today. I'm not sure that has anything to do with Weather, their whole notification priority system is borked.)
I think -- and I might be wrong, since this is from over a decade ago -- that when I first used Dark Sky, I ended up disabling notifications because it would constantly warn me of precipitation, but then when I checked the graph there was none because the model had since updated, and I wound up turning them off. So notification thresholds are probably something hard to get right, and what is appropriate for one geographic area might not be optimal for another.
I've sort of transitioned to using Ventusky and Windy to checkout the big picture stuff, then I make up my own mind about precipitation. I live in the PNW of the US and our terrain is so varied that forecasting services are kind of meh in general. They're decent for "it might rain for a while today" but anything hyperlocal tends to get bad because of the terrain in Oregon.
Also I really like a tool called Forecast Advisor. https://www.forecastadvisor.com/ . It shows you the accuracy of various forecasting services for your area.
I use it whenever I travel. I don't stick with one forecast site because depending on the terrain/location their accuracy changes drastically.
Certain models are better for certain geographical features depending on the location. I tend to hangout around a lot of mountains and the difference in forecast models makes a huge difference.
The Doppler radar that "live" precipitation comes from takes 4-6 min to complete a scan, and then obviously it takes a few minutes for that all to be ingested, update models, and push to devices.
The "live" weather from Apple (and when it was Dark Sky) has always been a prediction from about 10 min ago. And if it's raining where you are but dry six blocks to the north (as happens all the time), it's understandable why it gets it wrong.
Lived in the same general area (just outside a major metropolitan area) where I use DarkSky and now Apple Weather app.
DarkSky has better data vis and more reliable prediction. Apple Weather consistently over predicts snow fall amounts and many times I’ve had to use the Feedback to correct it on current conditions (e.g. raining when it says no rain or vice versa). I believe DarkSky had the same feedback feature but I never needed it this much.
Most of the time AW is fine but it’s less good to the point I’ve considered alternatives.
Apple Weather is nothing like this.
That said, I'd still bet a dollar (that to be fair, I might lose) that Apple today is less accurate, and if they're just as accurate twelve years on, that's a fail as well.
In all seriousness I heard some good things of dark sky. My current weather app is windy.com and I believe it's more built for surfers and such (??) - not sure what the best android weather app is.
Scrolling through the Dark Sky screenshots, I can recognize many of the same things now incorporated with Apple’s. And Apple does offer location specific notifications of rain which I find to be pretty accurate, about as accurate as Dark Sky.
There’s largely a perception problem with Apple. People loved Dark Sky as an independent small app that worked well, before Apple took it and destroyed it. Now, even if Apple incorporated all of the same data and features, it still wouldn’t give that same spark of joy people had.
This is what I really liked about DarkSky. I didn’t have to read and understand the forecast, I could simply glance at it and intuitively have an understanding of the day’s weather. Apple lost this, and I think it is what gave DarkSky so much value.
Even without any text labels, you should be able to get a feel for what the weather is and how it will change:
- Hourly plots like Dark Sky, with everything (temperature, rain, AQI, weather conditions) in a single plot.
- The change in temperature visualized with both color and space. Space is obvious (higher -> hotter); color ranges from red for hottest to blue for coldest. All the visible plots share the same color-temperature mapping. So the gradient block to the left shows both the temperature range for that day as well as how it compares to other days.
- Finally, there is a weekly overview at the top.
Huge bag of data for you to mess around with. I've started to use it to do my own weather forecasting instead of relying on forecasting services. Where I live has a radar gap(Oregon) and ridiculously varied terrain, so forecasts aren't great anyway.
Interesting, I think it's gotten worse over time. Even basics like what the temperature will be in a few days. It's consistently ~5+ degrees off on the low side.
Not a replacement at all for Android subscribers!
The developer is very responsive, lots of UI customization (both app and widgets) is possible, and pricing is reasonable.
Still the best of all the weather apps though.
I should be able to release it before the end of January.
Best, Tomas
Both give you a huge amount of layers and datasets to mess around with. Windy recently changed their radar stuff, though, so it might be a bit confusing.
https://github.com/PranshulGG/WeatherMaster
All the other android apps mentioned here have the same issue.
I might try to open an issue in their GH, or even a PR... A toggle for "denser graphs" and a setting for hourly resolution could do wonders.
1: https://github.com/davidtakac/bura/
It really sucks. Do they use the same data? I've noticed Apple Weather is substantially less accurate than Dark Sky. If Dark Sky told me it was going to rain in 10 minutes for 7 minutes, that's what was going to happen. If Apple Weather says it, well, maybe.
Apple says they use data from the weather channel, but this varies based on country. It used to say right in the app, but it seems like they removed that in favor of this link:
https://support.apple.com/en-us/105038
Then one day it just stopped working and it all went away. But it took apple ~18 months or so to kill it off, if I remember right.
But there was never any refund or whatnot.
I've used it daily since.
[1] https://news.ycombinator.com/item?id=34155191
Windy.app is for wind based water activities. Windy.com is a data-heavy weather information site.
Nice things:
There's a screen shot showing what it used to look like before gradients: https://github.com/Leftium/weather-sense
- Toggling C/F also toggles the scale on the radar to km. Eventually, I will get around to adding a dedicated settings page.
- However, the app was designed so one could get a sense of the weather without numeric labels: temperature is a very relative experience, so use the spatial/color cues to compare yesterday, today, and forecast days.
- Notice now much more space C needs when toggling between C and F. F's 0 to 99 range fits the natural range of temperatures humans experience (weather, body temperature). Humans just don't experience anything beyond 50 degrees C. At the same time, a single delta C is too large for the precision human bodies can detect. (Humans need something closer to 0.5C precision, which is what 1 degree F is.)
- As as result C needs nearly twice as much horizontal space compared to F: due to going negative more often and needing an extra decimal for minimal precision required.
- Now most gradients are disabled by default. (toggle here: https://weather-sense.leftium.com/wmo-codes)
- Also added shadow/glow to plot lines so they "pop out" more.
I'm not sure which parts you think are blue and teal. Open to suggestions for better colors! (There are only so many colors, and I like keeping the precipitation related colors all bluish.)
Some differences:
- Shows weather from yesterday for comparison
- All hourly plot trackers connected; not just the top one
- Includes AQI
- Sky color visualization (try scrubbing across dawn/dusk!)
- Non-precipitation colors approximate sky color (haziness)
- Temperature variation visualized both spatially and with colors
- Data source is Open Meteo
- Planned: 60 minutely forecast like https://openweathermap.org
- https://weather-sense.leftium.com/?n=nyc
n is short for "name" and uses the Open Meteo geocoding API[1].
[1]: https://open-meteo.com/en/docs/geocoding-api
* Adopt a colour scheme with similarity to the old BOM?
* Some way to store longer baseline movie animations in local state so people can avoid cost in you but run the weather radar for longer?
* Tide info? Hyper specific to people who do water things. Willyweather does this really well.
I use Willyweather and Windy. I used to use a weather app written by some mob called "shifty jelly" and their git logs were .. hysterical. Drunk fairy penguins seemed to cause most of the bugs.
- Location search & display: no English version, only localized names, which is hard to reason if you don't know native language.
- Summary: a) what does precip. in % mean and how it correlates with cm? b) pressure is only displayed in hPa, while some countries prefer mmHg; c) what does ozone index mean and why is it important?
- Next 24 hours: visual indication for temp variation through the day would be easier to reason than just by looking at numbers (as discussed in TFA).
- Next 7 days: a) unnecessary precision for y-axis (e.g. 10.9°C vs. 11°C); b) band overshoot actual values (e.g. if I see 10..−10°C, I assume that would be max/min temp, but in fact it is 8..−9°C, which is impossible to tell without hovering mouse over); c) no horiz. line through 0°C; d) no horiz. lines through y-axis ticks, which makes it harder to reason about values closer to the end of the graph; e) precip. in cm tells little, especially when band is alike (0.00..0.80 cm) - peaks on graphs look like a lot, in fact they are not? g) seeing blue precip. graph subconsciously means 'rain' to me, while in fact it would be snow; f) labels for y-axis are at the same time very small, rotated 90° and also take too much horizontal space from the graph.
- Map: moving mouse over next 7 days graph causes time shown on map to change that would make sense if map's timeline would cover all 7 days, but it only covers small part of today.
- Week: a) fog icons look like they have solid white square background, which seems to be off compared to other icons; b) low/high values are hard to reason about, especially when it says 'Low … at 11am' and there is no tick labeled '11am' (10am .. 12pm) - displaying a line through coldest/warmest hours with °C value next to it would be much easier to understand.
Also: displaying air quality prediction based on last year's AQ would be helpful.
I have learned to ignore its predictions. It will say that it's sunny outside, and I'll look out the window, and we're having a hailstorm.
I've learned that I just want to look at the radar. There's a big difference between "it's going to drizzle all day" and "spotty storms within 25 miles of you"
[1]: https://polarhabits.com/mobile
- The data is from https://open-meteo.com
- It would be trivial to connect the historical weather API (back to 1940): https://openmeteo.substack.com/p/processing-90-tb-historical...
Polaroid
Pebble
Palm
Oldsmobile
Tower Records
Borders
Pan Am
https://repebble.com/
Whenever I travel I find it pretty helpful. Certain services are just garbage in some areas.
For example, Foreca is like 84% accurate for my home location, but it's only 60% accurate for one of the cabins I frequent.
Weather APIs are pretty open. What's stopping you?
Check out their compare feature. Brilliant.
Eg: Seattle vs London https://weatherspark.com/compare/y/913~45062/Comparison-of-t...
(Not affiliated. Just an admirer.)
In all other places in the app, the low is to the left of high.
The idea that Apple is full of good designers should be forgotten. They're as mid and sloppy as any other large tech company.
During a cold snap one cold night will show up as the low for two consecutive days instead of a single “overnight”
This topic raises an issue I’ve had in mind for a while, companies are not realizing their true value when they sell out to some exit, as is evident by the fact that the companies Andy what they created end up being taken out behind the shed. If a competitor is willing to pay a certain amount without extreme pain to the point of convulsion or you don’t get air tight contract that prevents killing off the service/product without it remitting back to the founders or being made open source, you are being low-balled.
Taking the Wunderground example, those folks would have ended up owning the weather channel and probably buying or merging with Dark Sky and being the data provider to Apple instead of the Weather Channel characters (in case you don’t know about that entity) owning and killing off their baby.
they've been lobbying for like a decade to get NOAA defunded. They're basically the Intuit/turbotax of the meteorology world.
One of the things that I've seen with them that I haven't seen with others is the cloud cover by layer.
https://www.yr.no/en/details/graph/2-6301678/United%20States...
https://www.yr.no/en/details/table/2-6301678/United%20States...
For doing photography (sunsets) there's a significant difference between 50% high clouds and 50% low clouds.
A year and a half or something later.. I recently started a project of my own trying to bring all "weather dependent" photo opportunities together in one place, if you wouldn't mind I would be happy to experiment with bringing Sunsethue data to https://photoweather.app - your prediction model is certainly a lot more sophisticated than mine and it would be very cool to offer that
https://kachelmannwetter.com/ has data from dozens of models, but only in separate maps I think.
It was a rare example at the time when it was _the_ webapp better than any existing 'native' apps.
So it would be like "60% chance of rain after 2pm, total amount less than 1/10th of an inch"
Look at image. Scroll down to find the next image button. Scroll back up to look at image. On desktop
I now use Weathergraph which does it differently but I would go back to Dark Sky (and pay for it) in a flash.
It shows the correct things and on a phone understands that showing the temperatures across the screen is useless as if I go out I want to know what the weather is like when I might make the journey back in 8+ hours time. I might not care what the weather is in 4 hours time as I will be inside.
A website dedicated to data visualization and it's totally broken on Desktop Firefox. If they had just created a straightforward article, it would be perfectly legible, but all the flashy-flash just makes it unintelligible.
I really disliked that change - was that just me?
The table of user "context and situation" is a great document. You can easily envision authoring this table and scrolling to the right of your initial columns (A,B) to see further into the design process,
A) "When I hear about a storm, I want to prepare my loved ones, my property, etc.
B) Storm forecast ... : - Where is the storm right now and is it heading my direction?
[...]
N) _Show the storm front using _directional arrows_ ... (compact and replaces need for animation)_
The last section concludes in praise of the design and includes this: _"rigorously iterated on data visualization design". I wish we would have seen evidence of this, principally in the form of older screen shots of the design.
I think design iteration is the difference between mere good design and good products, and legendary product design.
Personally, I'd love to see a write up of my favorite whipping post, Transit App. Oh boy did that app go down hill, and with such great potential.
22 more comments available on Hacker News