End of content perfectionism
For decades, live performance required massive infrastructure. Musicians needed stages, lighting rigs, sound engineers. TV hosts needed studios, cameras, production crews. Quality meant budget, and budget meant barriers. The best productions felt cinematic because they cost cinematic money.
That’s collapsing. AI is democratizing production across software, music, and video, and while costs are plummeting and access is exploding, we’re also drowning in an ocean of perfectly polished noise. What matters now isn’t production value. It’s taste, perspective, and human perception. The twist is that both how media gets produced and how audiences perceive it are shifting simultaneously.
The Sora moment
When I started writing this, Sora was still in research preview. By the time I’m finishing, it’s launched publicly and made a massive splash. The quality is exceptional, people are having fun with it, but what’s more interesting is the “Sora or Real” trend that emerged: people posting videos asking others to guess whether they’re watching reality or AI generation.
Extrapolate these model capabilities forward, and we’re not far from stitching thousands of Sora clips into long-form content or generating entire videos that look so real you can’t distinguish them from footage shot on a camera. That future is measured in months, not years.
The shift to raw
Creators are abandoning the studio. They’re going live, shooting on phones, doing spontaneous pop-ups with three days’ notice. This is happening because humans crave connection with other humans, and as AI-generated content becomes indistinguishable from human work, we’re paying obsessive attention to what makes that connection possible. We’re hunting for imperfections, constantly gauging whether what we’re seeing is real or generated.
Anyone with a trained eye can spot AI now. Long dashes are the tell, bullet points with bolded prefixes are another. Creators who want human connection have to make content that’s unmistakably human, and the market is forcing this shift whether they’re ready or not.
What’s a bit sad is that when I work so much with language models, I start to pick up traits they produce. My somewhat original way of writing is converging with how the models write, and it takes active effort not to do that.
What human actually means
Human is imperfect. It’s not having the best camera. It’s shaky footage, no makeup, visible mistakes. But mostly, it’s being present. Pre-recording is easy because you can edit, alter, polish until it’s pristine. Showing up live and performing without a safety net is hard, which is exactly why it works.
There’s something else happening too: nostalgia aesthetics are making a comeback. VHS grain, CRT monitors, early internet vibes. This isn’t accidental. When everything can look perfect, looking deliberately imperfect or vintage becomes a signal of authenticity. It says “a human made this choice.”
Live performance and spontaneous content are becoming the future of digital brands. More AI video means more pressure to look human. Livestreams are harder to fake. iPhone shots with shaky cameras turn imperfection into a feature. (Apple only shoots on iPhone, Theo told me recently, though they use cinema-grade lenses.)
The playbook here is to lean into what makes humans different: we’re unpredictable, spontaneous, full of surprises. Short announcement windows, pop-ups, ad-hoc events. You don’t have to take yourself so seriously when you’re just showing up as yourself.
Who’s already winning
Look at OpenAI’s livestreams. Their GPT-4 and GPT-4o announcements have more views than any of their polished videos. Fred Again sells out shows with three days’ notice, and this isn’t random. It’s engineered. He sends audio files through WhatsApp because he “likes the compressed sound,” which is a deliberate choice to make humanity a feature rather than a bug.
TBPN is taking sports (the world’s largest media fuel) and turning it into tech content through livestreaming, doing trading cards that put humans at the center while robots and AI compete with them. IRL streaming is racking up millions of views, and every YC company is putting out high-production cinematic videos that nobody remembers. If your content feels overproduced, it probably is.
For decades, live performance needed massive infrastructure. Musicians: stages, lighting rigs, sound engineers. TV hosts: studios, cameras, crews. Quality meant budget, budget meant barriers. Best productions felt cinematic because they cost cinematic money.
That collapses. AI democratizes production in software, music, video. Costs plummet, access explodes, and we drown in perfectly polished noise. Production value no longer matters. Taste, perspective, human perception do. Twist: production and audience perception shift at same time.
The Sora moment
When I started this post, Sora was in research preview. As I finish, it launched publicly, made massive splash. Quality exceptional, people have fun. More interesting: “Sora or Real” trend. People post videos, others guess reality or AI.
Extrapolate: thousands of Sora clips stitched into long-form content, or whole videos indistinguishable from camera footage. Months away, not years.
The shift to raw
Creators abandon studio. Going live, shooting on phones, pop-ups with three days’ notice. Humans crave connection with humans. As AI content becomes indistinguishable from human work, we obsess over what makes that connection possible. We hunt imperfections, constantly gauge real or generated.
Trained eye spots AI now. Long dashes are tell. Bullet points with bolded prefixes, another. Creators wanting human connection must make unmistakably human content. Market forces it, ready or not.
Bit sad: working so much with language models, I pick up their traits. My somewhat original writing converges with model writing. Resisting takes active effort.
What human actually means
Human is imperfect. Not best camera. Shaky footage, no makeup, visible mistakes. Mostly: being present. Pre-recording is easy: edit, alter, polish to pristine. Live, no safety net, is hard. Exactly why it works.
Also: nostalgia aesthetics return. VHS grain, CRT monitors, early internet vibes. Not accidental. When everything can look perfect, deliberate imperfection or vintage signals authenticity. “A human made this choice.”
Live and spontaneous content become future of digital brands. More AI video, more pressure to look human. Livestreams harder to fake. Shaky iPhone shots turn imperfection into feature. (Apple shoots only on iPhone, Theo told me recently, with cinema-grade lenses.)
Playbook: lean into human difference. Unpredictable, spontaneous, surprising. Short announcement windows, pop-ups, ad-hoc events. Show up as yourself, less seriousness required.
Who’s already winning
OpenAI livestreams. GPT-4 and GPT-4o announcements out-view every polished video of theirs. Fred Again sells out shows on three days’ notice. Not random, engineered. He sends audio through WhatsApp because he “likes the compressed sound.” Humanity as feature, not bug.
TBPN takes sports (world’s largest media fuel), makes it tech content via livestream, trading cards with humans at center while robots and AI compete with them. IRL streaming racks up millions of views. Every YC company ships cinematic videos nobody remembers. Content feels overproduced? It probably is.