Ideas Weave Every Narrative with AI.
The true AI threat isn’t a dramatic sci-fi extinction event, but the mundane, immediate devaluation of human labor based merely on employer belief in AI’s competence. Before the technology even works properly, your bargaining power evaporates. We are sleepwalking into a world where we are economically useless and politically invisible, making structural safety nets like UBI politically impossible.
Everyone from Alipay to Taobao is slapping chatbots onto their homepages. But most of these ‘AI assistants’ are pseudo-needs that just add a typing tax to your life. The real value of AI isn’t answering questions—it’s executing tasks. And platforms are terrified of doing that because it threatens their ad revenue.
WeChat’s recent visually jarring input box update sparked massive backlash for being ‘ugly.’ But this aggressive UI nudge isn’t a design failure—it’s a calculated strategy to shift users from typing to AI-assisted voice interaction. The future isn’t a better keyboard; it’s an environment where speaking is the default.
The AI industry is obsessed with finding the single ‘best’ frontier model, but the real battle is shifting to the routing layer. If you’re building on a single provider, you aren’t innovating—you’re handing your moat to a future competitor. The real threat isn’t Chinese open-source models; it’s the frontier labs holding your data.
Baidu just absorbed its massive AI coding assistant into a universal desktop agent. This isn’t a product merger—it’s a ruthless land grab. The era of standalone AI tools is dead, and the war to become your single ‘super-agent’ entry point has just begun.
The AI jobs boom isn’t a recovery—it’s a construction crew. Companies are hiring to implement AI initiatives, but many explicitly plan to reduce headcount once systems are running. The more successful these jobs are, the faster they make themselves obsolete. The apocalypse wasn’t postponed; it was outsourced to the people building it.
AI products are falling into a trap of high usage but zero retention. DAU and token volume are vanity metrics. If unplugging your AI doesn’t stop a customer’s business workflow, you have no pricing power. To win renewals, you must embed your product into critical business paths, turn middle managers from threatened obstacles into AI supervisors, and shift from selling efficiency to selling operational dependency.
I’ve spent two years building AI digital employees for enterprise clients. But when I tried to build one for my own business, I choked. The bottleneck isn’t AI capability—it’s the unexamined messiness of your own proprietary processes. Adding more agents doesn’t scale productivity; it multiplies your QA overhead. The optimal strategy is regressing to a single, highly constrained task with a strict human handoff protocol.
Programming has always been trapped between art and plumbing. For decades, commercial reality forced us to kill our inner perfectionist and ship pragmatic, ugly code just to make a living. But as AI automates that exact utilitarian drudgery, human developers will be forced to become pure artists—or face obsolescence.
You spend hours organizing AI skill files to feel indispensable, but you are actually building the manual for your own replacement. Skill files are a fragile, temporary bridge over AI’s ignorance. As model capabilities expand, your perfectly curated directories will vanish into the latent space. The real competitive advantage isn’t the skill itself, but the human judgment to know what is still worth encoding—and when to delete it.