← Back to home Notes[AI-assisted]

The Creator Flywheel: Notes on the Creation Problem

Turning viewers into creators, then growing creators into top creators: one flywheel, two engines, and three levers for thinking about where modeling fits in the creation problem on a short-video platform.

The ideas here are my own, from a hand-drawn sketch; the write-up was AI-assisted.

In one line: the distribution system powers the “distribution loop”; content/topic modeling and inspiration modeling power “creator growth”. They are the flywheel’s two engines.

1. The flywheel

flowchart LR
    C["Viewer"] -->|"break the ice"| F["First-time creator"]
    F -->|"grow"| T["Top creator"]
    T -->|"produce"| H["Top content / breakouts"]
    H -->|"distribute back"| C
    H -.->|"inspire"| F
    H -.->|"inspire"| T

Four nodes, one closed loop:

StepConversionEngine
Viewer → first-time creatorBreak the ice, build a creation habitCreation-propensity modeling (entry)
First-time creator → top creatorSettle on topics, a niche and a fan base; keep growingContent/topic modeling · inspiration modeling (growth)
Top creator → top content / breakoutsProduce top content and breakouts—
Top content → viewersGood content is distributed back, closing the loopDistribution & search system (consumption side)

The two dashed arrows are the inspire loop: top content and breakouts in turn spark first-time and top creators to create more actively. It is not part of the linear main loop; it is a “re-creation” feedback that runs through the whole flywheel.

2. The two engines

flowchart TB
    subgraph EA["Engine A · creator growth (creation side)"]
        direction TB
        A1["Content / topic modeling"] --> A2["Inspiration modeling"] --> A3["First-time creator → top creator"]
    end
    subgraph EB["Engine B · distribution loop (consumption side)"]
        direction TB
        B1["Distribution & search system"] --> B2["Top content / breakouts sent back to viewers"] --> B3["Loop closes: more viewing, more new creators"]
    end
  • Engine A (growth): helps first-time creators find topics, gives them inspiration, and pushes them toward the top. It decides how fast the flywheel spins.
  • Engine B (distribution): gets good content to viewers and pulls in new creators. It decides whether the flywheel spins at all.

3. Three levers for growing creators

This is Engine A in detail. For content/topic modeling and inspiration modeling to actually help creators, it comes down to the three things below, and much of it is delivered through the distribution and search system.

flowchart LR
    CR["Creator"] --> I["Inspiration"]
    CR --> Q["Quick feedback"]
    CR --> R["Role models"]
    I --> I1["Related topics"]
    I --> I2["Better content on the same topic"]
    I --> I3["New information"]
    Q --> Q1["Fast likes / comments / follows<br/>after cold-start exposure"]
    R --> R1["Find bigger top creators"]
    R --> R2["Learn from role models to build a profile"]
LeverWhat it meansMainly delivered by
InspirationRelated topics, better content on the same topic, new informationDistribution & search / inspiration modeling / content modeling
Quick feedbackGetting likes, comments and follows quickly during cold startCold-start distribution
Role modelsFinding bigger top creators and learning from them to build a profileCreator discovery

4. Where each modeling task sits in the flywheel

TaskStepIn one line
Creation-propensity modeling (will this viewer try creating; will they remix or imitate what they are watching)Viewer → first-time creatorTurn a viewer into a first-time creator (entry conversion)
Inspiration modelingFirst-time creator → top creatorGive inspiration, drive growth, help creators move up
Distribution & search systemTop content → viewers, plus the three leversDistribution loop + cold-start feedback + inspiration and role-model discovery
Creator EcosystemCreationRecommender Systems
← All writing阅读中文版 →