A Solo Dev's 44-Day AI Mini-Program Postmortem: 36 Carts, 8 Payments, and a 15% Checkout Rate
For solo developers shipping AI-assisted products, this is a concrete case where AI speed amplifies the bottleneck of human attention, not just build speed. The failures here - no funnel instrumentation, dashing five features at once, mistaking ad fetch counts for real exposure - are generic to any indie builder, not unique to WeChat. West.
The creator of ChuCaiJi, a WeChat mini-program for home cooking and ordering, shipped 57 commits in 44 days and made ¥375.50 total. But the retrospective's sharp edge is the membership funnel: 36 orders created, only 8 paid, a 22.2% success rate that drops to 15% in the last 12 days. The author admits the payment funnel was never instrumented, so the abandonment reasons are unknown.
Traffic plateaued around 100 daily UVs after a holiday spike, ad revenue sawtoothed with one eCPM spike never repeated, and ad impressions only rendered 37.9% of fetched requests. The developer's key lesson: AI compressed build time dramatically, but five simultaneous feature launches in one week consumed attention faster than it could be managed, letting the funnel rot unnoticed for 12 days.
AI product development speed becomes a liability when the operator's attention is the limiting factor; five features in one week meant none of them got the follow-up measurement they needed.
A funnel with no instrumentation is a promise to fix later that never gets prioritized until the numbers get bad enough to force it.
Counting ad fetches as exposure is misleading: 37.9% impression rate means over half of ad requests never render, so the real ad inventory is far smaller than the raw fetch number suggests.
The pattern of small-tier orders after a period of growth suggests a large portion of checkout traffic is exploratory, not committed; the product is not yet converting first-time visitors into believers.