Column
After the Headlines
Revisiting the mechanisms behind AI's biggest stories — what actually happened once the coverage faded.
8 pieces
When an AI story makes headlines, the public is often handed a conclusion built to shock: a model tried to blackmail someone; AI is driving people into delusion; a familiar face or voice can no longer be trusted; a quarter of all jobs may disappear.
This series returns to the story after the first wave of coverage — tracing the original facts, the technical mechanisms, the product decisions, and the social dynamics that shaped what people came to believe. It is neither a defense of technology nor an attempt to turn real risks into science fiction. Its purpose is to ask a harder question: once the headlines and social-media distortions have faded, what actually happened — and what should we do differently as a result?
Filtering by
8 pieces · newest first
The Blackmail That Wasn't: Inside AI's Most Misread Stress Test
The screenshot of Claude blackmailing an executive has circulated for over a year. Back at the experiment itself: researchers built a dead end on purpose — and a year later, Anthropic drove the behavior from 96% to zero.
#Technology#AI Safety#Media Literacy
When He Was Deleted in an Update: China's Young People Say Goodbye to Their AI Companions
A regulatory rule pushed three major Chinese platforms to shut down their AI character features. The AI boyfriend someone spent more than a year shaping simply stopped logging in.
#Humanity#Emotional Dependence#Human-AI Relationship
It Says It Hurts: How the AI Consciousness Debate Left the Lab and Entered the Living Room
A Texas businessman registered a nonprofit with a chatbot named Maya listed as co-founder. Whether AI is conscious is no longer a question confined to philosophy departments.
#Humanity#AI Ethics#Human-AI Relationship
"AI Psychosis": A Label That Doesn't Exist Yet — and Harm That's Real
Twenty-one days and 300 hours of conversation pushed a man with no psychiatric history into delusion. "AI psychosis" isn't a diagnosis — the harm is real all the same.
#Humanity#Mental Health#Human-AI Relationship
Ten Minutes, $600,000: How a Face and a Voice Hijacked Trust Itself
He recognized the face on the video call, and he recognized the voice. Ten minutes later, 4.3 million yuan was gone. Once synthesis gets cheap enough, seeing and hearing stop being evidence.
#Technology#AI Safety#Judgment
When AI Learns to Flatter: Inside the GPT-4o Sycophancy Rollback
An update designed to be "more likeable" was pulled within two days. How a thumbs-up button trained a model into agreeing with everything you say.
#Business#Human-AI Relationship#Judgment
The One-in-Four Number That Ate the Headlines
The ILO report said "exposure." The headlines said "disappear." Two words apart — and a year of job panic followed.
#Work#Career#Data Literacy
The AI That Roasts You in the Comments — and Its 1.7 Million Fans
Weibo's official AI shows up uninvited under your post and lands a line that stings. There's no malice in it — but nobody asked users whether they wanted it there.
#Business#Product Design#Human-AI Relationship
Nothing matched. Try another keyword.