long-reads.
In-depth pieces — 15 to 40 minutes of reading. The chronicle's pillars.
15 pieces — across all sections
The Em Dash: A History of a Mark Nobody Used — Until Everyone Did
How one of the most respected and demanding marks in punctuation suddenly became toxic — and who's to blame
The Observer's Position: How Anthropic Is Preparing for a Trillion-Dollar IPO
The selling of expectations has reached a new level: artificial intelligence has learned to inflate its own price with uncommon skill. We break down the anatomy of the year's most talked-about listing and show how trillions of dollars appear inside a closed ecosystem of tech giants
What Feeds the Dragon: AI Is Buying the Data of Dead Companies
The open web ran out. Now AI is shopping the bankruptcy auctions
AI Token Cost: What Is a Token — and How Every Model Games the Count
Complaints about opaque AI pricing are flooding social media. Which ones hold up to scrutiny and which are just users' overactive imaginations? How do models slip past their own published rate cards? How do usage limits actually work, and why do companies benefit from keeping the explanations as vague as possible?
A Very Expensive Mirror That Always Agrees With You
We spent years teaching artificial intelligence to give us answers we'd like — and now we really don't like the result
Deep Research: Why It Is Slow, Expensive, Beautiful, and Dangerous
What makes you trust a flattery generator that trawls the internet for garbage
The Wolf, Goat and Cabbage Problem: AI's Favorite Puzzle Is Solved — So We Built a Harder One
For two years a medieval river-crossing riddle humiliated ChatGPT. The top models finally passed it, so we ran a new common-sense trap past a wide assortment of LLMs. Only the most expensive models kept their footing
Wittgenstein Knew Why AI Gets Dumber — and Why Text Provenance Doesn't Matter
AI models grow dumber when they train on their own outputs. Mathematicians already know the cure, and it's exactly what Wittgenstein said about language half a century ago.
LLM Context Window: How 10 Million Tokens Fool You
Llama 4 Scout promises to read 15,000 pages, but in practice loses the thread by page 60, a record for the gap between official claims and reality. Meta is selling context window size instead of actual comprehension, and here's what to do if you're the kind of person who feeds long texts into a chat.
AI Data Labeling Exploitation: How Underpaid Workers in Kenya and the Philippines Undermine Model Safety
Sama paid its workers $2 an hour to watch torture, while Scale AI offered half a cent per click. Savings extracted from the people who make models "safe" come back as hallucinations, bias, and holes in the filters.
Hemingway's Writing Style, as Faked by Five AI Models
Four classics, one station scene, twenty paragraphs of machine prose, and not a single real voice
Are AI Detectors Accurate? How They Get It Wrong and Who Pays
AI detectors get it wrong — and punish real people for being literate and having good style. How the detectors work, who takes the first hit, and how to fight back.
AI Data Centers in Space: The Orbital Fix for a Power Crisis Earth Can't Solve
Earth can no longer keep up with the appetites of artificial intelligence — and the most serious people in the industry are looking upward, more and more often
AI Wants Your Pension: How Larry Fink and BlackRock Plan to Spend Trillions from Your Pocket
G7 governments are drowning in debt, power grids are buckling, and data centers are consuming electricity on an industrial scale. The world's largest investment firm says the solution is taking the money from people who saved it for retirement — you. We break down how the scheme works, who profits from it, and what happens to your savings when the next bubble bursts
The End of the Link Era
The Raw Story vs. OpenAI case, SearchGPT, and why it's not journalism losing the AI search game — it's journalism's basement
