10 days into the cohort, roughly 20% of 50 students have crossed the monetization threshold—that's the scorecard Juejin author "Xiao Fei Chang" delivered this week. He's taking the Xiaohongshu virtual-products side hustle and breaking it into an AI "Skill" pipeline (a single-task AI workflow module). What we want to know: is this method actually replicable, and what does that 20% really mean?
What this is
Author "Xiao Fei Chang" bills himself as a former programmer now running a one-person company. He operates a Xiaohongshu virtual-products training camp that teaches cohort members to use AI to produce virtual goods (perler-bead patterns, wallpapers, etc.) and sell them on the platform. The core play is decomposing the full pipeline into discrete Skills: a Selection Skill that screens promising categories, a Full-stack Note Skill that writes notes and benchmarks competitors, and a Perler-Bead Pattern Skill that generates the pixel-art patterns themselves.
The latest case in point is perler-bead patterns: the user inputs text or a photo, and AI automatically handles subject detection → background removal → pixelation → color-number matching, outputting a printable pattern complete with grid, color codes, pixel preview, and an SVG file. The Skills live inside a "Workbench" interface that cohort members invoke directly.
Industry view
The numbers look shiny, but they need a cool head:
Sample bias. The author's self-reported 20% is based on roughly 50 paid participants; lurkers and the coaching team have already been filtered out. The training camp itself is on sale, so that figure carries marketing motive.
Category fragility. Perler-bead patterns are essentially "compute + API" material re-processing with fuzzy copyright boundaries. Xiaohongshu has tightened policies on virtual-products categories in recent years; a niche that works today may get traffic-throttled tomorrow.
Moat problem. The Skill pipeline carries a low engineering bar—anyone can replicate a similar workflow, and the competitive floor flattens fast.
But the underlying trend is real. AI is pushing social-commerce content-production barriers close to zero, and the traditional "designer + ops + customer service" small-team setup is starting to look bloated.
Impact on regular people
For enterprise IT: The traditional small-company default of "one designer + one ops" is starting to look redundant; one-person teams that string AI workflows together will become a more economical competitive option.
For individual careers: Side-hustle trial-and-error costs have genuinely dropped to a historic low, but a 20% short-term monetization rate is not stable income. More precisely: everyone can get started, but only a minority actually makes it through.
For the consumer market: Virtual-product categories will keep proliferating, with widening quality divergence. Consumers will need to get used to picking the real goods out of an ever-larger sea of AI-generated content.