On 10 July 2024 Adam Mosseri, who runs Instagram, posted a Reel to say Instagram would not be doing long-form video. TikTok was testing hour-long uploads at the time and everyone wanted to know whether Instagram would follow. No, he said, and he gave a reason that was better than most product reasons. Instagram exists to connect you with your friends. A ten- or twenty-minute video means you see less of your friends and send them less, and nobody forwards a twenty-minute video to a mate. Short form could do that job. Long form couldn't. It was, he said, part of Instagram's core identity, and he wasn't going to risk it.
On 26 October 2026, last Monday, Mosseri and Instagram's product chief Tessa Lyons posted another video. Instagram now has a format called Features: horizontal videos up to 90 minutes long, living on your profile, with up to eight Reels attached to each one as trailers. The app that killed IGTV in 2022 and promised in 2024 that it wouldn't go past a few minutes has built a format four and a half times longer than the one it refused.
Ninety minutes, from the company that said no to twenty.
I don't think Mosseri was wrong in 2024. I think he was right about what short form does, and he's just found out what it can't.
Thirty-six of each
Seven years before Features, Instagram ran the cleanest experiment anyone's ever run on what short form can't do, and it ran it by accident.
In May 2019 an influencer called Arii launched a clothing line. She had 2.6 million followers on Instagram, more people than live in Auckland and Wellington put together. The line was called Era, it had seven products, and her manufacturer wanted 36 orders of each before they'd run the machines. Call it 250 pre-orders. Her launch was two posts.
Thirteen days later she posted that production was cancelled. People who'd said they would buy didn't. She called it a wake-up call, deleted the post, and the internet reached its verdict in an afternoon: followers aren't customers. The bubble is bursting. It's been the verdict ever since, every time a creator brand folds, and it has the shape of all the best wrong answers, which is that it's true and it explains nothing.
Because here's what nobody asked. What did those two posts give 2.6 million people to do? They'd been shown a shirt. They had not been shown why this shirt, why her, why now, or why the shirt in the photo didn't match the feed they'd followed her for, which several of them pointed out. The format had no room for any of it. Reels at the time were capped at fifteen seconds, feed posts were a picture and a caption, and the whole machine was built to make you feel something before your thumb moved.
She had 2.6 million people in the doorway. Nobody buys anything standing in a doorway.
The missing premise
Aristotle noticed the problem 2,300 years before Instagram built it. A good speaker, he wrote in the Rhetoric, leaves out the premise the audience already holds, and the hearer supplies it himself. He had a name for the argument with the gap in it, the enthymeme, and he thought it was the most persuasive thing a speaker could do, for a reason that's almost rude in its practicality. A conclusion someone reaches themselves is one they'll defend. A conclusion you hand them is one they'll pick apart.
Pascal said the same thing in the 1600s. Psychologists have been proving it since the 1960s under names like self-persuasion and the elaboration likelihood model. And in 2012 three behavioural economists proved it with furniture.
Norton, Mochon and Ariely found that people who assembled an IKEA box themselves valued it 63% higher than people handed the same box already built. The IKEA effect, and everyone's heard of it. The second finding is the one nobody quotes. When the build was interrupted and left incomplete, the effect vanished. Not reduced. Gone. No premium at all for a thing you weren't allowed to finish.
That's the whole essay in one experiment. Persuasion is a gap you leave for the audience to fill, and an argument they complete becomes theirs. An argument they can't complete, because it's fifteen seconds long and the premise they'd supply has nowhere to go, earns nothing. It's a half-built box. You can show it to 2.6 million people and sell 36 of nothing.
Short form can't leave a gap. It can only leave things out.
The difference matters. Leaving a gap is a choice about what the audience will supply. Leaving things out is what happens when there's no room. A fifteen-second clip doesn't trust you to finish the argument. It never had the space to start one.
Why the platform that said no built the room
Mosseri's 2024 reasoning was that long video doesn't get shared between friends, and he was right. It doesn't. What he's since discovered is that the thing it does instead is the thing Instagram was failing to sell.
Look at the shape of Features. The 90-minute video isn't the thing you're served. The Reels are. Each Feature carries up to eight short clips, and the clips are described, in Instagram's own framing, as the way in. The short form is the trailer. The long form is the film. Instagram has drawn, in its product architecture, the exact distinction Mosseri said it didn't need: short form is for being found, and something else is for being stayed with.
It didn't arrive at this alone. Every platform built to be short has made the same journey, and all of them in the last two years.
TikTok launched at fifteen seconds and now takes hour-long uploads, and its creator payout programme won't share ad revenue on anything under a minute. YouTube Shorts was capped at sixty seconds until October 2024, when it went to three minutes. Twitter was 140 characters for a decade, then 280, then 25,000 for anyone paying, then a 100,000-character Articles product. Instagram's own Reels went from 90 seconds to three minutes in January 2025 because creators said 90 seconds was too short, and the cap kept climbing until last Monday, when it stopped being a Reel at all.
They'll let you be short. They won't pay you to be short.
The pattern is the tell. Every one of these companies has reached the same conclusion from the same data: attention is the free tier and duration is the product. They'll hand reach to anyone who can cause a spike. They keep the money for people who can hold a room, because that's where the advertisers are, because that's where the belief is.
The creators noticed before the platforms did. The most common complaint on creator forums for the last two years is some version of "my Shorts get reach and zero subscribers." A newsletter covering the Features launch put it in one line: short form gets people interested, long form keeps them. Another was less kind. Features, it said, is good for Meta and good for brands, and creators just get another platform to feed.
Both are right. What Instagram built last Monday is a room. Whether it's your room, or Meta's, is a separate question, and we'll get to it.
What attention actually costs
The objection waiting at this point is the one in every workshop deck: nobody watches past eight seconds. Attention spans are collapsing. A goldfish holds nine.
The goldfish number comes from a 2015 Microsoft Canada marketing report, which cited a commercial statistics website, which cited nothing. The BBC went looking for the study and found there wasn't one. The researcher who has actually measured this, Gloria Mark at UC Irvine, has spent twenty years logging how long people stay on one screen before switching: two and a half minutes in 2004, 75 seconds in 2012, 47 seconds now. But she measures switching, not capacity. Nobody's brain got smaller. The cost of leaving dropped to zero.
So attention isn't collapsing. It's cheap. It costs the viewer nothing to give you three seconds and nothing to take them back, and a feed can hand you a million of those three-second gifts before lunch. What a feed can't hand you is the ten minutes in which a person decides you're worth believing, because that isn't something a person can be served. It's something they choose to give, and they give it, in 2026, to the Joe Rogan Experience at two hours thirty-nine minutes an episode, to a four-thousand-page book series, to a three-hour superhero film that just retook the all-time box office record. Not because anyone's attention span grew. Because that's where the believing happens, and people are still doing it.
Mosseri didn't build Features because attention came back. He built it because attention was never the thing he was short of.
Two objections, one of them new
The strongest case against this comes from Byron Sharp, and for most of the economy he's right. Buyers don't need to believe anything; most purchases are low-involvement choices from a repertoire, and advertising works by refreshing memory so you're the brand that comes to mind in the aisle. Reach everyone, be distinctive, stay short, because a fifteen-second spot gets you most of what a thirty does.
For toothpaste, correct. Nobody completes an argument about toothpaste. But Sharp's laws are written for brands with mass reach, and nobody buys a mortgage, a magnet, a kitchen or a consultant from the repertoire. They buy after a decision, and a decision is a gap somebody has to fill. Sharp's own researchers concede it from the other side: short ads need simple, straight-line stories, and a straight line has no gap in it. So the claim is narrow. Anything considered, anything with a reputation on the line, needs the room. For toothpaste, carry on.
The second objection is newer, and it's the one Features will have to answer. Length is now free. An AI will write you ten thousand words before your coffee's cold, and it'll script a 90-minute video too. The platforms know it: Substack shipped an AI detector in July, LinkedIn added a "seems like AI slop" button nine days later and people pressed it a million times in three weeks. If a long piece used to prove a human spent time, it doesn't any more.
Agreed, and look at what follows. Length was never the proof. Length was the venue for the proof. What a viewer was always checking, in hour two of a podcast or page four of an essay, was whether a specific accountable person could hold up under duration. Nobody can keep a persona going for ninety minutes. That's the whole reason people watch ninety minutes. The test still works; it's just that the person now has to be visibly there, with a name and a face and something to lose, because the words on their own no longer carry it.
Which is also the answer to whose room Features is. Meta built it, Meta will sell the ads in it, and Meta won't, as far as anyone can tell, be paying creators for it. But the thing that happens in the room, a person being believed, belongs to the person. That part was never Meta's to own.
What Arii would have had
Run 2019 again with last Monday's Instagram.
She'd still have the 2.6 million, and they'd still have been shown a shirt in a Reel. But the Reel would have been a trailer, and behind it would have sat forty minutes of her: where the fabric came from, why the cut, the shirt on her at the end of a day, the critic in the comments who said it didn't look like her feed, answered to camera. The missing premise, she wears this, therefore I could, would have had somewhere to be supplied or somewhere to be refused. Either way the argument would have been finished, by them, and they'd have defended whatever they decided.
Maybe she still sells 36 of nothing. But she'd have found out why, and the 250 who did buy would have been hers.
Short form is where you get discovered. Long form is where you get believed. Instagram took twenty-seven months and one public reversal to admit it, and it's now the only platform of its size that has built the distinction into the product. The trailer and the film. The doorway and the room.
So build the long thing. The hour, the essay, the conversation that runs past the point where you could still be performing. Put your name on it and leave the premise out. Your audience will finish it, and they'll defend the ending, because it's theirs.
Assumptions, stated plainly
I've made a few and I'd rather list them than have them found.
- That Instagram built Features for the reason I've given. Instagram's stated reasons are creator demand and its TV app. Mine is an inference from the shape of the product. The shape is strong evidence; it isn't a quote.
- That length is necessary, not sufficient. Most long content is padding. A 90-minute Feature can be 90 minutes of nothing. The claim is that belief needs the room, not that the room produces belief.
- That the trend is universal. It isn't. Netflix seasons have gone from 9.5 episodes to 8.3 since 2018, for cost reasons, and that's a real counterexample.
- That anyone has directly compared long and short on trust. As far as I can find, nobody has, with numbers. The IKEA study, the persuasion psychology and the podcast data are the nearest evidence, and I've leaned on them.
- That this is about considered purchases. It is. If you sell something people buy without thinking, Sharp is your man and this essay has nothing for you.
I'd like to be wrong about some of the above, particularly point 1. Tell me where it breaks.