Why You Hate Your Weather App
In March, we expect mercurial weather—intrat leo, exeunt agnus—but this March has taken things to an extreme. In Washington, D.C., where I live, the weather was eighty-four degrees and sunny one day, then just above freezing and snowing the next. Then came a promised torrential downpour, followed by tornado warnings that sent locals searching for suitable hideouts, though, in the end, the city got only a wan drizzle. In that latter scenario, and many others of late, I have heard grumblings about weather apps and have experienced their frustrations myself. Check the weather in the morning and you might see sunny icons; come afternoon, it’s raining. Wind gusts aren’t necessarily visible on Apple’s weather app unless you happen to scroll way down the page. A colleague wore open-toed shoes the other day because her phone listed a high of fifty-four degrees—but the balmiest temperature occurred around 2 A.M., and the rest of the day was frigid. Is the problem climate change, which is making the weather more and more difficult to predict? Is it artificial intelligence, which is increasingly integrated into weather forecasting, and often falls short? Is it just the general enshittification of all software?
The answer is all of the above. Judging from the volume and tenor of user complaints, weather might be second only to social media as a space in need of fresh disruption. One entrepreneur, Adam Grossman, is on his second attempt at building a better weather app. In 2010, inspired by his experience getting caught in a surprise downpour during a road trip to Cleveland, he co-founded Dark Sky, which specialized in real-time updates on inclement weather. The app, which cost $3.99, became such a hit in Apple’s App Store that Apple acquired Dark Sky in 2020, and integrated some of the app’s features into its own weather app, employing Grossman along the way. But Apple shut down Dark Sky in 2023, prompting an online outcry, and Grossman eventually left out of frustration with Apple’s sluggish corporate schedule of annual software updates. He planned to get out of the weather business and launch a new startup. But the temptation of trying again proved too great. Grossman told me recently, “Everything from the tech stack you use, to the techniques that you use, to how you visualize the data—it evolved over time.” But, he added, “We would make changes to Dark Sky, and people would get mad.” With Acme, a new app that he launched in February, he could start fresh.
Weather apps have a tendency to alienate their user bases, perhaps because people’s physical experiences—their plans, their dress, their commutes—so directly depend on an accurate report. As Jonas Downey, the co-founder of an app called Hello Weather, told me, “It doesn’t take much for the app to lose trust with someone.” One unforeseen storm might send users looking for an alternative. Ever since Dark Sky shut down, Downey added, “There’s been this absence in the market.” Acme (which has a subscription-based model, at twenty-five dollars a year) is less cluttered than Dark Sky by design. When you open the app, you are presented with the weather “Right Now,” the forecast for the “Next 24 Hours,” and the forecast for the “Next 10 Days”; each is listed under a bold text banner that resembles a print newspaper headline. The twenty-four-hour temperature forecast is shown as a fluctuating black line tagged with icons denoting the weather every three hours. Crucially, though, there are alternate forecasts that appear in the form of paler gray lines; sometimes the gray lines hew closely to the main black one; other times they stray considerably, indicating that the predictions are less reliable during that window. As Grossman put it to me, “Climate change is causing an increase in uncertainty. It sucks that we can’t predict the weather perfectly, but knowing that uncertainty is very useful.” He cited, as an example, one recent day in his home state of Connecticut, when the forecast predicted snow in the morning and rain in the evening; in between, the conditions were harder to pinpoint. Acme dealt with this by displaying alternate forecast lines for both rain and snow, coded in different colors—a visualization that doubled as an admission that the forecast that day was a bit of a mess.
Apple’s Weather, by contrast, seems designed to telegraph an aggressive certainty, which can contribute to incidents like my colleague’s shoe mishap. Its home screen’s hourly forecast consists of temperatures, weather icons, and illustrative background visuals, such as gloomy rain clouds, that sometimes fail to match the view out the window. The individual data points don’t seem to visually or intellectually knit together. Acme integrates prose descriptions—“Clear skies this afternoon, turning partly cloudy this evening”—and precipitation percentages into its own home screen, for what Grossman called “context”—the kind of narrative information that a television weather forecaster might provide. Words are better at suggesting ambiguity than plain numbers or emojis. Acme’s main and alternate forecast lines turn the weather into a story arc.
The differences among weather apps are largely a matter of presentation; most apps run on the same set of data and predictions available to anyone. In the United States, the National Oceanic and Atmospheric Administration (NOAA) and the National Weather Service release their data to the public for free. Businesses can also license data from the likes of the Weather Company, AccuWeather, or the European Centre for Medium-Range Weather Forecasts. Most indie weather apps rely on predictions from one data service or another. Brian Mueller, the creator and developer of a popular independent weather app called Carrot, told me that he spends in the six figures annually for an assortment of data sets and for server costs. “No one weather data source can ever be accurate for everybody all over the whole planet,” Mueller said. One of Acme’s advantages over its smaller competitors is that it generates its own, bespoke forecasts, comparing and merging the various data sources and predictions using machine learning, and adjusting for factors such as regional microclimates. The app also uses its own algorithm to determine which information to show a user at a given moment. A radar map will pop up at the top of the scroll, for instance, when inclement weather is predicted.
The challenge of weather-app creation lies both in the improbability of accurately predicting the weather and in the difficulty of designing something that works for any user, anywhere. “Everybody wants their own weather app,” Mueller told me. An Angeleno may care more about air quality, for instance, whereas a Bostonian wants to know the chance of snow. Carrot’s imperfect solution is to allow users to customize their own display, choosing which information to foreground, against a backdrop of chaotic animations and snarky jokes (“The temperature is low, but my disdain for you is even lower”). Hello Weather separates various stats—on UV or wind—into separate onscreen tiles. Acme’s answer, the most elegant of the three, is to show a minimum of information based on what matters most in a given moment. Downey, of Hello Weather, told me that the public’s faith in weather forecasts hasn’t been helped by the current political environment, as President Trump has worked to defund the institutions responsible for reliable scientific measurement, including with cuts to NOAA. “People started trusting the weather data less,” Downey said, leading to even more animosity toward their apps. Acme imparts a sense of trust by foregrounding its own lack of omniscience. As I write, I can see on the app that it’s unlikely to rain this afternoon—but that there’s an alternate forecast suggesting a possibility of snow. At least I know that I can’t be sure. ♦