AI 与技术

Evidence does not yet show an AI jobs apocalypse. But hiring data from the U.S. and 41 countries suggest a quieter shift at the entry level — one that could change how workers gain experience and move up the career ladder.

Internet regulation is shifting beyond harmful content toward recommender systems, defaults, addictive features and AI companions. The change could make online harms easier to regulate — but it also creates new risks for speech and privacy.

Generative AI has not yet produced a measurable white-collar wage collapse. This essay argues that the earlier pressure will fall on the paid first draft: the routine, standardised output that used to justify junior hours and freelance fees.

AI compute is already allocated through price, contracts, infrastructure ownership and public programs. The harder question is what happens when capacity becomes genuinely scarce.

AI can raise productivity while quietly changing the human skills, manual processes and organizational memory a company retains. What happens when the technology becomes unavailable?

Google is stronger than ever as a business and an AI company. The harder question is whether years of product shutdowns, a changing Search experience and early AI missteps have made long-term user trust harder to earn.

AI’s energy problem is usually framed as a question of electricity demand. The deeper risk is what happens if society reorganizes around always-available machine intelligence before it learns how to function when that capacity becomes scarce.

AI can help beginners or shut them out. Employers cutting junior hiring must explain how they will train the experienced workers they still need.

中美顶级模型的能力差距正在缩小，但资本、算力、创业密度、工业落地与人才环境仍然不对称。下一阶段AI竞争，不只看大模型榜单，而看谁能把能力变成持续生产率。

从反复不满意 AI 写作，到把事实、信号、假设、边界和禁止项写进软件合同，再到五篇真实文章 E2E 测试、63 项回归测试、MIT 开源和 GitHub v0.1.0 Release。