Anthropic just watermarked every word Claude writes. Copywriters are celebrating. I argue that's premature. The real story isn't detection. It's that we've built a culture where writing takes minutes and reading still takes hours.
The Zen of Adult Friendship Maintenance
Most of us believe org chaos protects our jobs. The reasoning goes: processes are messy, decisions are undocumented, judgment lives in people's heads — so AI can't touch what it can't even see clearly. I used to believe this too. Until I ran a small experiment that made me question it. Turns out, AI didn't need my process. It just needed to see my output.
Shad Linga: How to grasp a text’s intent without asking its author
A scholar at Ramakrishna Math told me about a method used to interpret the Vedas. And it turned out to be the sharpest tool I've found for a much older question: what a text says vs. what its author claims it says.
Judgment Was Supposed to Be the Moat. The Bitter Lesson Says Otherwise.
Most of us believe org chaos protects our jobs. The reasoning goes: processes are messy, decisions are undocumented, judgment lives in people's heads — so AI can't touch what it can't even see clearly. I used to believe this too. Until I ran a small experiment that made me question it. Turns out, AI didn't need my process. It just needed to see my output.
The AI Tells You Can’t Edit Out
I recently submitted an essay as part of my AI policy course. The assessors noted that AI use wasn't prohibited, but had to be documented. It got me thinking: what exactly would they look for?
Wrote Too Well. AI Suspected.
Jamir Nazir won a Caribbean regional prize at the Commonwealth Short Story Prize. 7,806 entries. One winner from his region. Days later, online commentators ran his story through AI detection tools and declared it machine-written. Granta's own publisher said the judges may have "awarded a prize to an instance of AI plagiarism." Three detection tools were used. Three different verdicts. This is the new plagiarism charge. Easy to throw. Hard to disprove. Burden of proof falls on the writer. I've been working on this problem: collecting AI tells, reverse-engineering prompts, ghostwriting for people who need to prove their own voice on the page. Here's what I've learned about what actually separates human writing from machine writing. Hint: it's not the em dashes.
The Self That Survives the Layoff
A year ago I wrote about an engineer who lost a $150k job and was driving for DoorDash, living in an RV. Last week, out of nowhere, I looked him up. He's now Head of Engineering at a startup. I felt relieved. Not just for him, but for everyone wondering when their number comes up. Wrote about what his story made me think about: identity, layoffs, and the version of yourself that exists outside the job.
How AI Is Flattening Indian English Into American English
"No way he has written this!" Felt this way reading a LinkedIn post from a long-lost friend? He cannot write that well, you think. But that's below the belt. Maybe he improved. A steep learning curve, but not impossible. Yet something's amiss. Then it dawns. The language, the idioms, the metaphors. All highly polished. All... Continue Reading →
The Machine That Listens Too Well
It's a piece about how AI is affecting human ability to build and nurture relations with fellow-humans.
Translation Gap Between Product and Marketing
In an ideal scenario, product teams hold marketing accountable to technical truth. Marketing holds product teams accountable to the question the customer is actually asking. Neither dismisses the other.
Musings by Madhav 