Skip to content
Deepak Gupta markFuture Tech

Future Tech/consumer

Generative Entertainment Gives Everyone an Audience of One by 2032

On-demand fragmented the audience. Generative AI finishes the job by giving every person their own episode. By 2032 the shared cultural text is gone, because no two people watch the same thing at all.

// By 2032 · medium confidence · disruption 7/10

Prediction

// 2032

By 2032, mainstream streaming platforms will ship AI-generated, personalized entertainment, episodes and interactive narratives tuned to the individual, such that no two viewers watch the same thing.

Confidence██████████medium
Disruption██████████7/10

What dies

  • → appointment television
  • → cable tv set top boxes

Who wins

  • → Netflix
  • → YouTube
  • → OpenAI Sora

filed: 2026-06-14 · guptadeepak.com

The hook

Appointment television died because we stopped watching the same thing at the same time. Generative entertainment is the next step: we stop watching the same thing at all. When the algorithm can produce the episode, not just pick it, the audience does not fragment into groups. It fragments into individuals.

Thesis. On-demand split the audience into many small audiences watching at different times. Generative AI tuned to the individual splits it into an audience of one, and a culture with no common text loses its shared reference point entirely.

The story

The setup: personalization already ate the schedule

Streaming spent a decade personalizing everything except the content. The recommendation engine, the artwork, the autoplay queue, all of it bends to the individual. The one thing still held in common was the show itself: two people could at least watch the same season of the same series.

That last shared object is the next target. Once a model can generate a watchable episode on demand, there is no production reason to ship one version to everyone.

The hinge: generating content, not just choosing it

The shift is from recommendation to generation. Today the algorithm picks from a finite catalog. The hinge year is when it composes from an effectively infinite one: an episode pitched to your taste, a narrative that branches around your choices, a story with characters tuned to what keeps you watching.

At that point personalization stops being a layer on top of entertainment and becomes the entertainment. The output is unique to the viewer by construction.

Current state, 2026

Generative video is real but not yet a substitute for a studio season. Sora, Veo, and Runway produce convincing short scenes; coherence over a long, character-driven narrative is still hard, and cost per minute of high-quality generated video is still meaningful. [verify]

Interactive narrative has proven demand, from branching streaming experiments to the entire video-game medium. The missing piece is generating that branching content cheaply and on the fly rather than authoring every branch by hand.

Trajectory to 2032

Two curves carry this: model quality on long-form coherence, and inference cost. Both have moved fast, and entertainment is unusually tolerant of imperfection, because a slightly weird AI scene is a far lower-stakes error than a wrong answer in code or medicine.

By 2032 the likely shape is hybrid: human-authored worlds and characters as the frame, with AI generating the per-viewer episodes and branches inside them. Enough to make a meaningful share of viewing genuinely personal.

The holdouts

What resists is exactly what resisted on-demand: live, unspoilable events. You cannot personalize a real game in progress without breaking the thing that makes it matter. Sports, awards, and breaking news stay shared.

Prestige communal storytelling may also persist as a deliberate, premium counter-move, the same way vinyl and live theater survived their disruptions. But it becomes the exception people choose, not the default they inherit.

First signals (verify today)

Generative video crossed from novelty to usable in 2024 and 2025. OpenAI's Sora, Google's Veo, and Runway moved from short clips to longer, more controllable scenes, and YouTube began folding generative tools into creation. [verify] Streaming already personalizes everything around the content, the rows, the thumbnails, the recommendations, and watch time has shifted decisively from linear broadcast to on-demand. [verify] The remaining gap is generating the content itself per viewer, and that gap is closing as model quality and inference cost both improve.

Key data points

  • OpenAI's Sora, Google's Veo, and Runway advanced generative video from short clips toward longer, more controllable scenes across 2024 and 2025. [verify]
  • U.S. streaming watch time overtook broadcast and cable combined in the mid-2020s. [verify]
  • Cord-cutting continued accelerating through the 2020s, shrinking the linear, scheduled audience. [verify]
  • Streaming platforms already personalize recommendations, artwork, and queues per user, leaving the content itself as the last shared object. [verify]
  • Interactive and branching narrative has a proven mass audience via the video-game medium and streaming experiments. [verify]
  • Per-minute cost of high-quality generative video remains meaningful in 2026 but has fallen sharply year over year. [verify]
  • Live sports remain among the only content reliably drawing 100M-plus simultaneous U.S. viewers. [verify]

Contrarian angle

The comfortable framing is that AI just gives everyone more of what they love. The harder truth is what it removes: a common text. A culture coordinates partly through shared stories, the thing everyone saw and can reference. Generative entertainment optimizes each viewing for one person and, in aggregate, leaves a society with nothing it all watched. There is an ownership-to-access shift underneath. You once possessed a place in a shared audience, a moment held in common with strangers. Now you authenticate into a feed that generates a world for you alone. The most personalized entertainment in history may also be the loneliest, an audience of one by design.

The flip side

What this kills

The paired obituary in Tech Graveyard.

Read the obituary

FAQ

Won't this just be a gimmick like interactive TV experiments?

The earlier experiments hand-authored every branch, which capped them at a few paths. Generative models remove that ceiling by producing branches and episodes on demand, which is why this time the economics point toward scale rather than a one-off stunt.

Why is confidence only medium?

Long-form narrative coherence and per-minute generation cost are the two open variables. Both are improving, but it is genuinely uncertain whether they reach studio-substitute quality at consumer cost by 2032 rather than later in the decade.

Does this kill traditional shows entirely?

No. The likely outcome is hybrid: human-authored worlds and characters with AI-generated per-viewer episodes inside them, plus a surviving premium tier of deliberately communal, human-made storytelling.

What survives as shared culture?

Live, unspoilable events: sports, awards shows, election nights, breaking news. They resist personalization because their value depends on everyone watching the same thing in real time, which is also why appointment television clings to them.

Is this prediction a contrarian bet?

No. The direction is the consensus trajectory of personalization and generative video. The contrarian framing is only about the cost: that the price of an audience of one is the loss of a shared cultural text, which the optimistic version leaves out.

More from guptadeepak.com

Want the technical deep-dive behind this prediction?

Read the companion article

Related predictions

More from the consumer desk.