With new productivity, like AI, that doesn’t mean you can necessarily make money. Because you may still be trapped in the old production relations. Take the AI short-drama track, for example: AI is already strong enough, so why are so many people still not making money?
Over the past three years, I’ve spent a great deal of effort trying to break away from the existing production relations on the one hand, while on the other hand struggling to make a living and keep fighting within those same production relations. That’s probably the fate of ordinary people. Like in The Shawshank Redemption, no matter how disgusting or exhausting it gets in the end, you still have to crawl through that stretch of road.
AI short dramas have laid the problem bare: productivity has changed generation, but the power of distribution is still in the old hands.
In the first half of 2026, 221,900 new AI short dramas were released on Douyin, with peak output at about one title every 36 seconds. AI’s share once exceeded 95%. The rate of titles breaking 100 million views was only 0.47%. Based on a break-even line of about 50 million views, only about 1.3% could recoup their costs. AI animated dramas were even harsher: about one breakout hit per nearly 1,000 titles. Revenue per 10,000 views fell from 30-100 yuan in the second half of 2025 to 5-10 yuan, and some people said 11 dramas together only received 9.6 yuan in total. Production costs were cut by 90% on the surface, but traffic costs rose by more than 100%, and paid traffic often accounted for 70% of the entire chain. The result was that 90% of teams lost money, the number of production contractors was cut in half, and the ones making money were mainly those selling computing power, those selling courses, and the platforms themselves.
The money didn’t disappear; it was absorbed by the old structure. ByteDance used Seedance to bring down generation costs on one side, while using Douyin and Hongguo to keep distribution and settlement in its own hands on the other. A large portion of the production savings creators made was then spent back through paid traffic, returning to the same ad system. Non-transparent data, completion-rate metrics decided by the platform, canceled guarantees, lowered revenue-sharing coefficients — this isn’t a matter of technology not being strong enough, but of the residual claim rights not changing. The platform still controls attention allocation, while creators are still paid by the piece plus a lottery ticket. After the explosion of supply, what is scarce has never been images; it is the right to be seen.
Blockchain was originally supposed to change exactly this layer: content rights confirmation, traceable contributions, revenue automatically split by contract, and fewer layers of commission between creators and audiences. IP on-chain, smart revenue sharing, and a community jointly owning a drama’s follow-up earnings could, in theory, put together a system where “generation is cheap, and distribution is even cheaper.”
Unfortunately, in reality, it never connected. Regulation killed off public circulation, the user experience never got past the creator’s threshold, and discovery and recommendation never truly became decentralized. In the end, the attention market still returned to a few super apps.
So AI has pushed production capacity to the limit, while production relations remain centralized traffic-based rent.
So it’s not that AI isn’t strong enough. It’s that once it becomes strong, who has the power to decide whether this drama is worth money, and how the money is split, has hardly changed at all. Productivity can change generation overnight; the power to distribute it changes much more slowly.