AI Is Devouring All Software

AI remakes software from asset to consumable

A new piece in Huxiu (虎嗅) argues the next tech tectonic shift is not just more software, but AI-generated software that arrives on demand — fast, cheap and “good enough.” The author, a decade‑veteran UX designer and founder of a 30‑person firm, frames the change as an update to Marc Andreessen’s 2011 maxim “Why Software Is Eating the World”: fifteen years on, software has been eaten by AI. It has been reported that tools such as Anthropic’s Claude Code and OpenAI’s Codex — and reportedly a growing crop of clones like “OpenClaw” — now let non‑programmers describe a product in plain language and receive a deployable, full‑stack application in return. The phenomenon is being called “vibe coding.”

What this means for SaaS and product moats

Why should Western readers care? Because the economics that built giants like Salesforce and Adobe — big codebases, high switching costs, long maintenance cycles — are under pressure. The Huxiu writer describes building dozens of bots in Feishu (飞书), the workplace app from Bytedance (字节跳动), to automate company workflows and tune news‑ranking “skills” with human feedback. Many formerly expensive, hard‑to‑replicate features can now be produced quickly by prompting an LLM and stitching outputs together. The implication is stark: will enterprise software remain a long‑lived asset or become a disposable consumable for most everyday use cases?

Geopolitics, limits and the next stage

This shift will not be the same everywhere. Export controls on advanced AI chips, sanctions and trade policy are already shaping who can run the largest models and at what cost. It has been reported that China is accelerating investment in domestic models and stacks to close gaps created by Western restrictions — which could produce two parallel ecosystems rather than a single global market. And there are limits: bespoke, safety‑critical systems, deep platform integrations, and regulatory compliance still require engineering discipline and ongoing maintenance. So yes, AI is devouring much of what we called software — but what replaces it will prize data control, governance and orchestration as much as code.