Weather, Worth Reading — The Case for an Editorial Weather App

You opened your weather app this morning. You looked at the temperature, you saw whether there was a raindrop icon, and you closed it. You did not actually read anything.

Almost nobody does. The average weather app interaction is under four seconds, and the average person checks several times a day for information they cannot remember by the afternoon. The numbers go in and the numbers go out and the day proceeds as if the forecast had never been consulted.

This is the design problem. Every weather app in the store has access to roughly the same atmospheric data, and every one of them presents it in roughly the same way: a number for the temperature, a small symbol for the conditions, a percentage for the chance of rain, a grid for the hours ahead. Template in, template out. Identical forecast, identical treatment, interchangeable words.

The Vesper argument is that the forecast is not the product. The forecast is a raw material. The product is what you do with it.

Most weather apps are not apps — they are tables

Open any of the major weather apps on your phone and you will see the same structure. A hero number. A secondary number. A row of hourly cells. A strip of days across the bottom. Radar if the app is trying. An ad bar if the app is monetized through data.

That structure is not a design choice. It is an absence of one. It is what happens when you take a CSV from a forecast provider and render it at 60 frames per second. The engineer did their job. The designer arranged the cells. Nobody wrote anything.

You can tell which apps are built this way because they all sound identical. "Partly cloudy. High of 68. Winds W at 5 to 10 mph. Chance of rain 10 percent." Swap the numbers, swap the city, the sentence is the same. That sentence was not written for you. It was not written for anyone. It is a field in a database rendered through a template.

An editorial weather app does the one thing that template can't do: it reads the forecast and decides what the day is actually about.

What "editorial" actually means in software

The word "editorial" gets thrown around loosely, and in software it usually means "hand-picked content" or "a curator chose this." That is not what it means here.

Editorial in the sense Vesper uses it is a discipline borrowed from magazines and newspapers. It means someone reads the raw inputs, decides what the dominant story is, writes a version of it in the voice of a specific publication, and cuts everything that does not serve that story. An editorial weather app applies that same discipline to the forecast for your specific city, on this specific day.

The inputs are the same ones every weather app has. Pressure. Temperature. Dew point. Wind direction and speed. Cloud cover. Precipitation probability. Sunrise and sunset. UV index. The editorial pass is what happens after.

The editorial pass asks three questions. What is the dominant story of this day? What is the reader about to decide? And what is the shortest possible sentence that answers both? Everything else is cut.

The five things a template cannot do

A template can render numbers in a grid. It cannot do the things that make a forecast worth reading.

Translate data into intent

A template says "68 and sunny." An editorial brief says "the kind of afternoon that makes you want to eat lunch outside." The number is the same. The decision the number should trigger is the point.

Hold a point of view across weather

A template is neutral. Every forecast sounds the same to every reader. An editorial brief has opinions — the writer thinks a marine layer morning is a gift, thinks January light in New England is underrated, thinks August in Houston is a challenge that deserves respect rather than defeat. The point of view is the reason to read.

Respect the reader's time

A template fills the screen. An editorial brief earns its space. If the day is nothing remarkable, the brief is two sentences. If the day is a pressure gradient event that will make the afternoon feel three degrees colder than the number suggests, the brief spends the extra sentence on that. Length is earned, not padded.

Adjust for season and place

A template treats 68 degrees as 68 degrees. An editorial writer knows that 68 in Denver in March is a celebration and 68 in Miami in November is a cold snap. The same data reads differently in different places, and a writer who knows the place translates accordingly.

Tell you what matters, not what's there

A template lists every available field. An editorial brief picks one or two. If the defining fact of the day is a wind shift at three o'clock, the brief is about the wind shift. If the defining fact is that the sky is going to do something beautiful at sunset, the brief is about the sunset. The cut is the craft.

The apps that tried to be editorial, and what happened

Vesper is not the first weather app to notice this problem. Several earlier apps pushed in the editorial direction and stopped short in interesting ways.

Dark Sky was editorial about design. The hyperlocal minute-by-minute precipitation forecast was a point of view rendered as a product. The problem was that Dark Sky was editorial about the chart and not about the words — the copy surrounding the chart was still templates underneath. Apple bought Dark Sky in 2020, integrated the precipitation technology into Apple Weather, and the words stayed templates. The design discipline lived on. The writing discipline never existed to begin with.

Carrot Weather is editorial about humor. Every forecast comes with a wisecrack in one of several selectable personality modes. This is editorial in the sense that someone wrote the jokes, but the jokes are canned, not responsive to the day. A Carrot forecast on a genuinely interesting morning reads the same as a Carrot forecast on a boring Tuesday, because the day-specific content is still template and only the personality wrapper is written. We respect Carrot. We are doing something different.

Apple Weather has the data, the design, and the restraint. What it does not have is voice. The copy is careful and inoffensive, which is the correct choice for a system app that ships to a billion users. But correct and inoffensive is exactly the opposite of editorial. The cost of shipping to a billion users is that you cannot have a point of view, and the cost of not having a point of view is that nobody reads you.

The window these three left open is the one Vesper is trying to fill. Editorial about the writing itself. Responsive to the specific day. Willing to have opinions.

Why Vesper is doing it this way

There is an assumption inside software that more data is always better. More refresh rate. More charts. More sensors. More machine learning. The forecast of the future, this argument goes, is the one with the most numbers per square inch of screen.

We think this is wrong and we think the data shows we are right. The average weather app session is three to five seconds and the average retention of what was displayed is approximately zero. The apps that win on data density win on the app store metric of "screen time per session" and lose on the metric of "did the reader learn anything."

Vesper optimizes for a different metric. Did you close the app knowing what the day is about? Did you make one decision differently because of what you read? Did you feel like the forecast was for you specifically, not for the generic category of "person in your zip code"?

That metric is harder to ship and impossible to template. It is the metric the rest of the industry is either ignoring or has given up on. We think it is the only one that matters.

A brief, demonstrated

To show what we mean, here is a brief generated for Portland on a real morning.

The inputs were ordinary. Temperature 52, low overnight of 46, dew point 49, pressure rising slowly from 30.08 to 30.12, cloud cover breaking from overcast to scattered by noon, wind west at 4 mph, sunrise at 6:21, sunset at 7:48, UV 4 mid-afternoon. A template would render this as "mostly cloudy, high of 58, low of 46, chance of rain 10%." Four seconds of reading, zero seconds of understanding.

The Vesper brief for that morning read:

Marine layer overnight, lifting by ten. The Willamette will give you back the light around noon and the afternoon will feel like the kind of clear Portland gives you three times a week but pretends is rare. Pack a jacket you can take off.

Thirty-seven words. The dominant story was the marine layer lifting. The decision was what to wear given that the morning and afternoon would feel meaningfully different. The specific sensory anchor was "the Willamette will give you back the light around noon" — something that is true of Portland specifically and not of any other city.

That brief took a human writer two minutes to revise from a first draft. It took a template zero seconds to generate the alternate version. The difference in those two minutes is what Vesper is selling.

What a Vesper brief is deliberately not

Part of the discipline of editorial weather is knowing what to leave out. A brief is not a joke, not an alert, not a data dump, and not a diary entry.

It is not a joke because a joke lands once and then gets in the way. An editorial brief wants to be read every morning, which means it cannot depend on novelty for its value. Carrot does jokes well. We are doing something the reader will want to return to for decades, not a gag that stales in a week.

It is not an alert because an alert is a command. Briefs do not command. They describe. If there is a severe weather event, the system sends an alert through the standard channel — that is the operating system's job. The brief is for normal days and unusual days alike, in a voice that stays consistent across both.

It is not a data dump because the whole point is to remove the data the reader did not need. A brief that lists six metrics has failed. A brief that leaves out five metrics to land the one that actually shapes the day has succeeded.

It is not a diary entry because the writer is not the subject. The reader is. The brief is written about the day from the reader's perspective, not about the writer's feelings or the writer's life. The voice has opinions but the opinions serve the reader's decision, not the writer's ego.

The Vesper take on weather writing

Weather is the most universally shared topic in human conversation. Everyone experiences it, everyone has opinions about it, everyone has felt their plans change because of it. Given how central weather is to daily life, it is genuinely strange that the category of software most people check three times a day has almost no voice.

We think the reason is historical and fixable. Weather apps started as utilities in an era when phone screens were small and data was expensive, which forced a minimalist data-first design language. That language calcified into a category convention. When Dark Sky introduced hyperlocal precipitation it was treated as a radical departure, when really it was only the first crack in a design pattern that deserved much more disruption.

The forecast is a starting point, not an ending point. The template apps stop at the starting point because stopping is cheap and nobody complains. Editorial weather apps walk the forecast all the way to a sentence the reader is willing to read. That walk is the whole product.

Read more about what a brief actually sounds like on our Vesper Brief feature page, or see the comparison with the major weather apps we built to explain what we are doing differently.

More editorial weather reading

The Vesper journal is a small library of editorial essays about weather, written in the same voice the briefs are written in. If this piece resonated, you may also want to read Why We Built Vesper, which explains the founding story, and UV Index Explained, which applies the editorial method to a specific and widely misunderstood number.

For city-specific editorial voice samples, the briefs section now hosts short editorial snippets for ten American cities, each demonstrating how the same voice adapts to the personality of a specific place.

A frequent question

People ask whether editorial weather can scale. Can a human writer produce a brief for every city every morning? The answer is that the writing discipline and the scale mechanism are separable problems. Vesper's briefs are generated by the same language model that writes for the rest of the software, operating under an editorial style guide that was written by people and refined through thousands of examples.

The point is not that a human sits at a keyboard at four in the morning composing briefs for every city on earth. The point is that the style guide, the cut discipline, and the voice rules were all written by people who care about the writing. The model is the mechanism. The editorial pass is the content.

Template weather apps are generated by models that were never given an editorial style guide. That is the difference, and it is the difference that produces different output.

Where this goes from here

The Vesper thesis is that the next generation of weather apps will be editorial, the way the last generation was hyperlocal and the generation before that was just digital. Editorial is the axis of improvement that is still wide open.

If we are right about this, then in five years the major weather apps will all have some version of a written brief, and the arguments will be about whose voice is best rather than whose radar refreshes fastest. That is the future we are trying to build toward.

If we are wrong, then people will keep checking weather apps for four seconds and closing them without reading anything, which is the present, and we will have produced a beautiful artifact for the small number of people who wanted to read their forecast instead of scan it. We are content with either outcome, because the people who wanted to read their forecast were never going to be served by the template apps and deserve a weather app that was built for them.

Weather, worth reading. That is the phrase. It is not a slogan. It is the design constraint. Every feature we build has to pass the test: does this make the forecast more worth reading, or does it just add another chart to a screen nobody looks at?

The answers to that question are shipping now, one brief at a time. Start with the Vesper Brief feature page, read a few of the city briefs, and see whether a forecast written to be read feels different from a forecast written to be glanced at.

It should. That is the whole point.