Chasing AI tools for a year, zero output: a serial entrepreneur's reflection

The trap of early adoption

A widely read essay republished by Huxiu’s TechFlow (深潮 TechFlow) argues that frantic early adoption of every new AI model and agent can become a productivity trap rather than an advantage. The author, a serial entrepreneur, says he spent months testing multiple programming assistants, image generators and agent frameworks — and delivered nothing. Is being “first” always a win? The piece warns that early signals feel like progress, but often they are merely dopamine-driven motion with no durable output.

It has been reported that many startups which hurried to build thin products on top of GPT‑3-like models have since folded, and that the current wave is already separating builders who use tools as means from tinkerers who treat tools as ends. In China this pressure is compounded by an intense local crawl for advantage: incumbent players such as Baidu (百度) and Alibaba (阿里巴巴) are pouring resources into AI, venture activity is feverish, and reportedly heightened geopolitics — export controls and US‑China tech competition — are pushing some teams to iterate faster on domestic stacks. That mix amplifies fear of being left behind.

Taste, attention, and the path forward

The essay’s core prescription is simple: attention is the scarcest resource. What separates winners from busybodies is “taste” — the judgment forged by consequence, not by consumption. The author points to Midjourney founder David Holz as an instructive case: instead of chasing every UI fad, Holz concentrated attention on the model and product core, even if that meant an odd interface. The lesson for Chinese founders and investors is the same as for Western ones: when tools become ubiquitous, differentiation comes from knowing what to build with them and, crucially, what to ignore.

So what should founders do next? Pick a single problem. Ship something. Let failure and consequence teach your taste. In an era when models, agents and compute are almost commoditized, the advantage lies less in tool‑tasting and more in sustained focus — one door, walked through. Which teams will turn early sight into lasting judgment? That question will shape the next wave of winners and losers in China’s fast‑moving AI ecosystem.