Seeing AIand the choices people make
Columns
From technology to business, from what happens to one person to what happens to a whole company — an ongoing record of the questions worth thinking hard about in the age of AI.
Filtering by
16 pieces · newest first
Your AI Just Changed 30 Files. Do You Dare Hit Accept?
People who don't write code are now getting agents to build things for them. The catch: an agent changes dozens of files at a time, and you need an undo button. That button is git. This piece makes the case for learning it, and comes with an interactive Git handbook — four places, four ideas, 25 commands, and prompts you can paste straight into your agent.
#Technology#Learning#AI Literacy#Workflow#Agents
Forget the One-Person Company: The Real Edge Belongs to Small Teams
The AI era's buzzword is the 'one-person company,' but the bigger shift is happening in small and mid-sized teams. A founder who spent years in big tech and now runs a 24-person startup says small-team productivity isn't built from efficiency tools — it's the product of low-loss communication, good atmosphere, and each person's drive. AI is only the amplifier.
#Business#Work#Human-AI Relationship#Management
A Plain-English Guide to Leading a Team of Agents
Someone spun up a six-agent team and burned through 40% of a week's quota on day one — three of the bots were re-checking the same pages. 'One agent works great, so a squad must work even better' — most of the time, it doesn't.
#Technology#Agents#Organization
A Plain-English Guide to Working With Agents
A dialog box pops up — 'Agent requests permission to modify 12 files. Allow / Deny' — and your hand is already on Allow. The real question was never how to talk to AI. It's how to work with something that can act on your behalf.
#Technology#Agents#Judgment
A Plain-English Guide to Agents That Can Actually Get Things Done
Same word, 'AI' — so why could last year's version only chat, while this year's takes on the whole job? What decides whether an AI can act usually isn't the chip itself, but the operating system built around it.
#Technology#Agents#AI Safety
A Plain-English Guide to Picking an AI Product
The thing you use every day isn't the model — it's the product. Understanding the six patches bolted onto every chat product is how you pick the right one out of a crowd that all call themselves intelligent.
#Technology#AI Literacy#Large Language Models
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
A Plain-English Guide to Talking to AI Well
Talking to an AI is nothing like using a search box. Four moves — Delegation, Description, Discernment, Diligence — turn a prompt into a working division of labour.
#Technology#AI Literacy#The 4D Framework
A Plain-English Guide to Large Language Models
An LLM's brilliance and its confident nonsense come from the same mechanism. Take the six steps apart and you can see exactly how a sentence gets generated.
#Technology#AI Literacy#Large Language Models
Nothing matched. Try another keyword.