🤓 Hollywood Can Now Move at the Speed of Culture
PLUS: Claude’s new model is here, and I’m already obsessed; Google wants its AI agents in your inbox, calendar, and shopping cart; Spotify is handing fans the AI remix button—and more
👉 A quick note before we get into it:
Lately, the conversations I’ve been having have gotten a lot more interesting.
I’ve been talking to more people in entertainment who are sitting on genuinely fascinating business problems and want to explore whether AI can help solve them in new ways.
The most promising of these have been with people who are deeply curious and open, willing to experiment, and have the authority and resources to do just that.
That combination is rare, and it’s leading to ideas and potential projects that I’m super excited to explore.
Things are also moving even faster on the technology side, and I need more space to keep building my own skills alongside the work.
So I’m shifting this newsletter off a fixed schedule. I’ll write when I have something worth your time—might be one week, might be three, might be longer.
To my paid subscribers: I’m also hoping this opens up room for me to put together some things just for you. Nothing to announce yet, but it’s on my mind..
In the meantime, I’ll be on LinkedIn sharing shorter takes and updates, so make sure you’re following me there.
Really looking forward to what’s ahead, and sharing more of it with you.
Now, on to this edition.
470 AI-generated short dramas a day.
That’s how many Chinese platforms were reportedly launching in January: vertical, mobile-first, one-to-two-minute melodramas built with generative AI.
Episodes run a couple minutes. The plots move at a sprint.
Someone gets betrayed, secretly rich, kidnapped, reincarnated, avenged, or forced into marriage before your coffee gets cold.
By February, more than 127,000 AI-generated short dramas were already in circulation.
This is not the future of all entertainment.
But it may be the first place AI entertainment makes real business sense.
China is using these tools to flood the market with more content.
The better question is what happens when they’re used for something more ambitious.
A Format Built for AI Before AI Arrived
Short dramas were already a machine.
They run on repeatable tropes, extreme emotional hooks, fast testing, cliffhanger ads, and direct monetization.
Viewers get pulled in through ads on TikTok, Instagram, Facebook, and YouTube.
A few episodes are free. Then they’re pushed into apps where they pay to keep watching.
It’s closer to performance marketing with plot twists than anything Hollywood would call a show.
AI fits because the format already rewards exactly what AI makes possible: more attempts, faster iteration, cheaper production, and a wider creative range.
But it also expands what the format can attempt.
Fantasy, dragons, mermaids, elaborate costumes and visual effects—everything that used to blow the budget is now viable on micro-budgets.
A live-action short drama that cost around $137,000 in China in 2024 can now reportedly be generated for $7,000 to $14,000.
MIT Technology Review reported that North American short dramas can now be made for 80–90% less, with timelines collapsing from months to weeks.
At that price, companies don’t need every show to be good.
They need enough cheap bets, enough paid traffic, and a few breakouts.
And audiences are already watching.
China’s short-drama market generated roughly $6.9 billion in 2024, surpassing the country’s annual box office for the first time.
Omdia projects the global micro-drama market will grow from $11 billion in 2025 to $14 billion by the end of 2026, with the U.S. becoming a major revenue market.
It would be easy to look at this category and write it off as cheap content.
That would be lazy.
Audiences already spend enormous amounts of time with content that doesn’t look anything like prestige film or TV: TikTok and YouTube videos, podcasts, Twitch streams, fan edits.
Our definition of “quality” has also shifted.
Sometimes quality means craft, polish, and cinematic spectacle.
Sometimes it means relevance, intimacy, utility, or hitting the exact mood someone showed up with.
People don’t open these apps expecting The Bear.
They want the hook. The payoff. The betrayal. The revenge. The next ridiculous turn.
Short dramas deliver exactly that.
And clearly, a lot of people want it.
More Shows, Fewer Hits
The current AI entertainment pitch goes something like this:
If production gets cheaper, you can make more.
If you make more, you can test more.
If you test more, you find more hits.
It sounds logical. It may even be partly true.
China is already testing that at scale, but the early results are more complicated than the pitch.
Of the 127,800 AI dramas in circulation by February, just 0.117% crossed 100 million views, and that breakout rate has reportedly been declining as production volume climbs.
The top AI drama reached roughly 1 billion views. The top live-action short drama hit 4.4 billion.
AI can lower the cost of making entertainment.
But it doesn’t lower the difficulty of making people care.
Once the cost barrier drops, the harder questions take center stage:
Can it hold attention?
Does it build emotional connection?
Can it build fandom?
Can it stand out once everyone else can make the same thing?
Those are the questions that don’t get easier when production gets cheaper.
And as these tools advance and spread, more people will learn the playbook and more content will flood the market.
But audiences still only have so many hours in the day.
The winners won’t be the ones who produce the most, but the ones who make something people actually choose.
Which raises a more interesting question: if the advantage isn’t volume, what do these tools actually make possible?
When Stories Can Catch the Moment
Social media creators have always had something that traditional entertainment couldn’t match: the ability to move at the speed of culture.
A creator on TikTok can turn a cultural moment into a viral video the same afternoon.
A meme can spread faster than any marketing campaign.
But richer forms of entertainment—scripted series, serialized stories, anything that requires real production—has never been able to do that.
Development and productions take months or years. By the time a show or movie reaches an audience, the cultural moment that sparked it may have already passed.
That’s always been treated as just how the process works.
What AI offers talented storytellers goes beyond cost savings.
It’s the ability to go from idea to produced content fast enough to stay connected to what audiences care about right now.
To build entertainment that feels alive to the moment—not because it’s disposable, but because it was made by someone who understands what resonates and can act on it.
That opens up possibilities we haven’t even begun to explore yet.
A side character from an existing franchise unexpectedly blows up. Fans clip them, meme them, write theories, and flood social media with “give them their own show.”
The studio owns the IP. The demand is obvious. But development and production of a spin-off still takes far longer than the obsession may last.
With AI tools, a small team with sharp creative instincts could test and iterate on a short-format concept while demand is at its peak.
Or take Amazon MGM’s Artificial, Luca Guadagnino’s upcoming film starring Andrew Garfield about Sam Altman’s firing and return to OpenAI.
The film has reportedly wrapped and is aiming for an early 2027 release.
Meanwhile, the Musk v. OpenAI trial just surfaced newly revealed texts, testimony, and internal details that add new dimensions to the story.
AI tools could make it possible to integrate the new developments into the film even as the story continues to unfold.
Great stories have always been able to transcend their moment. The best films and shows connect across years and decades. That doesn’t change.
What changes is that storytellers no longer have to choose between craft and responsiveness.
Between making something good and making something timely.
For the first time, the tools exist to do both.
China’s AI short dramas are the rough first version of this.
They’re using the speed for volume, hooks, and conversion—470 titles a day optimized for getting someone to tap “pay to continue.”
Put it in the hands of talented storytellers and worldbuilders with stronger creative instincts and a sharper read on culture, and what you get isn’t more content.
It’s new kinds of entertainment that weren’t possible before.
The first wave of AI entertainment looks cheap, weird, addictive, and everywhere.
What comes next will be made by people who can finally create at the speed of their own ideas.
📈 Work with Me
AI Advising and Consulting: Strategic guidance to build sustainable AI strategies and adapt to shifting audience behavior
AI Training: Practical training that helps teams use AI more strategically, effectively, and safely—so it becomes a trusted thought partner, not just a task assistant
What You Need to Know About AI This Week ⚡
Clickable links appear underlined in emails and in orange in the Substack app.
🔥Claude’s new model is here, and I’m already obsessed.
Anthropic launched Claude Opus 4.8, its most capable model yet.
The company describes the improvements as modest, but after testing it for several hours, I think they’re underselling it.
This model isn’t just smarter. It has a much deeper understanding of psychology and is far more emotionally intelligent.
It reads intent better, catching what you actually mean beyond what you say.
And unlike other agentic models released in the last few months, it’s patient enough to sit in complex problems with you and consider various perspectives instead of racing to an answer.
It’s also a noticeably better writer, and better at writing in your voice with the right setup and context.
👉 The catch: there’s a new control for how much “effort” or “thinking” Claude puts into a task. Higher effort means more thorough responses, but it takes longer and burns through your usage limits like I’ve never seen.
I ran two rounds of complex steps inside a longer chat and it ate up half my sessions’ usage.
So be intentional about how you use it.
Save it for the specific steps and tasks that need it, and go back and forth between Claude, ChatGPT, and Codex depending on the job or steps in a more complex project.
If you’ve read me for a while, you know it usually takes me days, sometimes weeks, to learn a new model’s strengths and blind spots.
After a few hours with Opus 4.8, I already know it’s my favorite model I’ve ever used.
I cannot wait to spend the weekend exploring every corner of it.
🔎 Spotting AI-generated images just got a little easier.
OpenAI is adopting Google’s SynthID watermarking for images made in ChatGPT, Codex, and other tools powered by OpenAI’s image model.
Google is also adding SynthID checks to Search and Chrome, so people can verify images from tools that use the system.
Since ChatGPT’s image generator and Google’s Nano Banana are two of the most popular AI image tools right now, that’s a meaningful step toward easier verification.
If more companies adopt it, this could become a real standard. But Midjourney, many Chinese models, and plenty of other image generators haven’t signed on yet.
🧩 Google is building AI agents into everything you already use.
Google announced a flood of AI updates at its developer conference, I/O.
I found the whole thing underwhelming, but the main takeaway was this: Google is building agents directly into the products people already use.
This matters because agents need context about your work and life to be useful.
And Google already has a lot of it: your inbox, calendar, docs, files, search history, YouTube behavior, shopping activity, and payment layer.
So instead of asking people to switch to a new AI tool or build complicated AI workflows from scratch, Google can put agents inside the places where people already work, search, plan, watch, and shop.
That gives the company a huge advantage with everyday users.
A few updates are worth paying attention to:
1️⃣ Google is making Search more conversational and action-oriented.
Its new “intelligent Search box” is the biggest redesign of Search in more than 25 years, and can handle longer, more specific questions and pull from text, images, videos and even files.
Search is also moving closer to a place where users can research, compare, and act in one flow.
So search is becoming less like a list of links and more like an AI-guided decision-making process.
As I’ve written before, Google’s AI mode can now create interactive experiences including custom dashboards, widgets, and mini-apps on the fly.
2️⃣ Google is building agents that keep working in the background.
Gemini Spark is a 24/7 personal agent that can work across Google Workspace, connected tools, and the web to handle recurring tasks while asking for approval before higher-risk actions like sending emails.
Daily Brief can scan your Gmail, calendar, and tasks overnight, then give you a prioritized briefing in the morning.
3️⃣ Shopping is becoming more agentic.
Google’s Universal Cart can track price drops, watch inventory, show price history, and flag problems before you buy like incompatible PC parts sitting in your cart.
This connects directly to the Shopify data showing AI-referred shoppers arrive more ready to buy. AI is moving closer to the moment of decision.
4️⃣ Video creation is becoming more like directing.
Google’s new Omni model can generate and edit video from text, audio, image, and video prompts. It can also transform a clip’s style or add characters while keeping the original movement intact.
You can give it your own video and keep iterating with your own ideas through conversation, instead of starting from scratch.
It can also create videos using your own voice and likeness through avatars.
5️⃣ YouTube is becoming searchable at the exact-moment level.
The new Ask YouTube feature lets users ask longer, more specific questions, get an AI-generated summary, and jump straight to the relevant part of a video.
Many of these capabilities and features are paid and/or rolling out “later this summer.”
I still prefer ChatGPT, Codex, Claude, and Claude Cowork to Google’s AI tools.
But if you already live inside Google’s ecosystem, Google’s advantage is obvious: its agents can start with the context you’ve already given it.
I also need to look more closely at the privacy tradeoffs, especially because… this is still Google and these tools rely on sensitive context from Gmail, Calendar, Docs, Search, shopping activity, and the rest of your digital life.
Related Posts:
🎬 Hollywood writers fought AI. Now some are training it to survive.
A revealing first-person WIRED essay follows a working TV writer who turns to AI training gigs after Hollywood work dries up.
It’s a look inside the strange, grinding world behind AI training, where projects disappear overnight, rules keep changing, and highly skilled people are reduced to chasing tasks inside systems built to make AI more capable.
📚 AI writing accusations are getting harder to prove—and harder to ignore.
A prestigious international short-story prize is under fire after readers accused several winners of submitting AI-generated fiction.
The writers deny it.
Detection tools flagged the work, but remain unreliable.
The foundation says it can’t revoke awards without stronger proof.
This is the messy new phase for publishers and prize committees: when writing looks and feels AI-generated, readers may lose trust long before anyone can prove how it was made.
🎵 Spotify is letting fans make AI covers and authors make AI audiobooks.
Spotify and Universal Music Group struck a deal to let Premium users create AI-generated covers and remixes from songs by participating artists and songwriters.
The tool will be a paid add-on, and Spotify says it is built around consent, credit, and compensation for the artists and songwriters who opt in. All users will be able to play the AI-created tracks.
Spotify is also launching an ElevenLabs-powered tool that lets self-published authors create and publish AI-narrated audiobooks without exclusive contracts. It starts as an invite-only beta in June in select markets.
📋 ChatGPT just made filling out forms easier.
Upload a blank form, use voice mode to talk through the information that needs to be added, and ChatGPT’s new image model can fill it in and send back a completed version.
In case you missed the last edition, you can find it 👇:
That's all for this week. See you next time.
Thoughts, feedback and questions are always welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi