It would be simple to take that as an attack on the people behind the movie, but it is not. I have never met anyone in this business who intended to produce something forgettable. The colorists, the second unit, the writer on pass eleven of a scene being rewritten around a locked release date, the VFX house burning through the night: immense skill is poured into work that, by design at the system level, leaves no mark.
That divide between input and output has nothing to do with talent.
The incentives do.
And before I continue, I need to raise my own hand, because an essay about a trap written by someone outside it is worth very little. I have a confession to make. I have written the headline before I wrote the piece. I have cut something shorter because I was hesitant the reader would drop off. I have done work I knew was disposable. Everything that follows is written from within the thing it describes.
What The System Actually Pays For
Attention is now priced second by second. Impressions are tallied, engagement is tracked, and eyeballs are quoted to the decimal.
Herbert Simon anticipated this in 1971 when he wrote that a wealth of information creates a poverty of attention. What has changed is the granularity. That poverty is now monitored continuously, per viewer, per second, and sent straight back into the machine that determines what gets made next.
If your business depends on keeping someone on screen long enough to show another ad or avoid another cancellation, you stop optimizing for the film that alters their life and start optimizing for the film that keeps them from leaving. A story that lingers for a month is excellent. A story that holds attention for ninety minutes and then auto-plays into the next one is lucrative. Only one of those will appear in a quarterly report.
The Data Gives It Away
The clearest proof I know comes from Upworthy. Between 2013 and 2015, it ran a randomized trial on nearly every article it published. Later, researchers examined more than twelve thousand of those experiments. Each extra negative word in a headline increased click-through by roughly 2.3 percent.
Two point three percent. In the mythology of clickbait, that sounds nearly negligible. Yet an advantage that small disappears in any single instance and becomes unstoppable in the aggregate, compounding across every headline, every day, in a system that tests nonstop and preserves whatever works.
Remember: Upworthy built its brand on positivity. Even there, the negativity effect appeared, inside the organization least ideologically ready to embrace it. Nobody sat down and decided that darker headlines were the future. The tests made that decision, and the staff adapted.
The company’s declared values lost to the instruments created to serve them, and nobody needed to be corrupt or foolish for that to happen. If you have ever been in a room where the data pointed one way and your instincts pointed another, and watched the data prevail, you already understand this story. It simply had a larger budget.
Three Roads, One Destination
It is easy to collapse the tentpole franchise, the AI content farms, and call them one disease. They are not the same mechanism, though, and the distinction matters.
Studio caution is risk management stretched years into the future. It is a two-hundred-million-dollar wager on a date three years away, protected by familiar characters and a proven tonal register, with no feedback loop faster than an earnings call. Nobody is testing the third act against a retention curve. The content farm operates in the opposite direction, optimizing after the fact, at almost no marginal cost, with results readable within hours.
What unites them is selection pressure. This is convergent evolution: sharks and dolphins arriving at the same shape from entirely different lineages because the same environment favored it. That environment is an economy that pays for retention instead of resonance, and it produces the same outline at both ends of the budget spectrum.
The third road runs through your own edit suite, which is why I am writing this now instead of three years ago. Generative tools have ceased to be a story about garbage on other people’s platforms and become a line item in your own budget. Previs, storyboards, the temp score, the pickup shot nobody wants to fly back for. Some of it works. Plenty of it still does not. But in every finance meeting, the conversation has already shifted, and the question is no longer whether the tool matches the person. It is whether the gap is worth the gap in price.
What Is Actually New
Remove the novelty, and the retention critique is old. Simon in 1971, and a long chain of people repeating the same basic argument afterward.
One feature of this moment has no precedent. The cost of making competent, professional-looking work has collapsed to almost nothing. The floor has risen. The ceiling has not moved.
Competence Is No Longer Scarce
If your career was built on being consistently good, as most working careers are, the ground beneath it has shifted in a way that no previous forty years of media criticism prepared anyone for. Being solid used to be a moat. Now it is the admission price, and the water is rising.
What remains scarce is narrower and harder to counterfeit: judgment about what is worth making at all; taste, which is only ever the residue of a particular person having paid attention to particular things for a long time; and the specific, the scene only you would write because of something that happened to you in 2009 that has never happened to anyone else in exactly that way.
A model can generate a competent version of nearly anything. It cannot generate the thing that could only come from you, because it has no you.
The Thing With No Telemetry
There is no metric for the line a viewer underlines and returns to four years later.
There is no dashboard column for the film that got someone through a hard winter.
There is no instrument for the moment a scene gives a person the word for something they have been carrying without a name.
Those experiences leave no data trail. A system that distributes money according to what it can observe will not just fail to reward them. It cannot detect that they happened at all. Whatever is invisible to the meter is, in money’s eyes, something that never occurred.
Good work has not stopped being made. What is suffering is the work that lands most deeply, which is precisely what the system is least able to notice, and the tools now flooding the market are very good at producing exactly what the system can see.
Follow The Meter
Difficult work is flourishing in particular places right now. A four-hour Hungarian film finds an audience that, in 1985, would never have known it existed. Letterboxd has drawn a generation into repertory theaters. Long-form podcasting created a mass audience for three-hour conversations, which is not the behavior of a species that has lost the capacity to focus. Substack pays essayists directly for the demanding work the ad model penalizes.
Notice the common thread. Substack is a subscription, paid before reading and indifferent to how long it takes. Letterboxd does not monetize dwell time. Podcast advertising is priced on completion and loyalty, not on the click.
Every place where demanding work is winning right now is a place where someone found a way to get paid outside the retention metric.
That gives you a practical test, and it is the most useful thing here. Ask any deal in front of you one question: does my payment depend on how this performs after delivery, or not? Subscriptions, grants, commissions with a fixed fee and no performance clause, straight work-for-hire with a real budget: these pay before consumption or regardless of it. Revenue shares on platform views, anything with a bonus tied to a dashboard: these pay afterward, and they will shape the work whether or not anyone means them to.
A financier whose return depends on completion rates will give different notes than one whose return depends on subscriptions. You will feel that difference in the third act. Knowing which side of the table you are sitting across from before the meeting is the same kind of professional knowledge as knowing the tax rebate in the territory where you are shooting.
Two things somebody should build
Since diagnosis is the easy half, I will put two mechanisms on the table. Both are flawed. Both are concrete.
The 90-Day Meter.
Creator compensation calculated on return visits over ninety days instead of 24-hour engagement. Do people come back to it. Do they bookmark it, cite it, send it onward. Payment arrives three months later, weighted toward delayed engagement. It measures residue rather than dwell time.
The Expiration Feed.
A platform in which everything has a fixed thirty-day shelf life and is then erased. No archive, no evergreen search traffic. The only way to keep an audience is to make something people will seek out before the clock expires. It injects scarcity into a system that has none.
The precedents are not encouraging. YouTube already did something like this: in October 2012 it shifted weighting toward watch time to stop rewarding deceptive thumbnails, and what followed was the padded ten-minute video and the ninety-second preamble. A better metric created a new pathology within two years, because whatever you measure becomes the thing that gets manufactured. Medium paid writers on reading time from 2019 and by early 2026 had moved to paying on new-member conversion, drifting back toward the funnel. And Snapchat already built the Expiration Feed. Ephemerality produced volume rather than care, under the entirely rational logic that if nothing survives, nothing has to be good.
So neither would work exactly as written. Both would at least fail in ways that teach you something, which is more than the current arrangement accomplishes. That one fails constantly and so predictably that nobody even bothers documenting the failures.
One Cup At A Time
The disposable cup did not prevail because anyone liked drinking from it. It won because it cost less to make, was easier to stack, and demanded nothing from the person holding it. By the time those advantages compounded, most of the places that once served drinks in glass had quietly stopped.
Nobody decided that. Everybody decided it, a thousand times, one cup at a time.
That is the trap, and also the only exit from it. A gradient made of a million tiny choices can only be altered by a million tiny choices, and the one in front of you is the next thing you agree to make and the terms you agree to make it under. That does not scale, and none of us can opt out of eating. But competence is about to stop being worth much, and the only thing left with any scarcity value is the work that could not have come from anyone else.
That is an unusually good time to stop making the safe version.
You are reading this, I notice, on a platform, beneath a headline I chose so you would click it. The gradient runs beneath all of us, all the time, and staying alert to which way it tilts is most of the job now.
Let's hear your thoughts in the comments below!
Got an idea for a post? Or have you collaborated with Stage 32 members to create a project? We'd love to hear about it. Email Ashley at blog@stage32.com and let's get your post published!
Please help support your fellow Stage 32ers by sharing this on social. Check out the social media buttons at the top to share on Instagram @stage32 Twitter @stage32 Facebook @stage32 and LinkedIn @stage-32
